Daily Papers

A robot following a multi-step instruction needs to act on the right objects in the right order, even as the scene changes. It must also maintain awareness of task progress and determine when to advance. To address these challenges, we introduce Orchestra, an agent harness that separates task planning and execution management from action generation. Orchestra decomposes a complete multimodal instruction into ordered subtasks, tracks completed steps, and updates the locations of relevant objects and destinations during execution. It maintains a shared context containing the active subtask, spatial grounding, and recent observations. An action policy uses this context to generate actions, while a separately trained progress evaluator estimates when to advance to the next subtask. To evaluate whether robots trained on atomic instructions and short trajectories can execute longer subtask sequences, we introduce a benchmark named RMBench-Compose. Orchestra achieves a mean success rate of 68.22% across nine RMBench tasks, outperforming Mem-0 by 26.22 percentage points. On RMBench-Compose, it achieves an Average Success Length (ASL) of 7.35 subtasks, 2.46 times that of HarnessVLA. Across four real-world multi-step tasks, it improves mean ASL by 25.0% over Orchestra using the base π_{0.5} policy. These results show that Orchestra helps robots carry out longer sequences of subtasks while preserving the requested subtask order and acting on the specified objects. Our code and project page are available at https://youpengwen.github.io/Orchestra.

Vision-Language-Action (VLA) models have shown substantial potential in real-world robotic manipulation. However, fine-tuning these models through supervised learning struggles to achieve robust performance due to limited, inconsistent demonstrations, especially in contact-rich environments. In this paper, we propose a reinforced fine-tuning approach for VLA models, named ConRFT, which consists of offline and online fine-tuning with a unified consistency-based training objective, to address these challenges. In the offline stage, our method integrates behavior cloning and Q-learning to effectively extract policy from a small set of demonstrations and stabilize value estimating. In the online stage, the VLA model is further fine-tuned via consistency policy, with human interventions to ensure safe exploration and high sample efficiency. We evaluate our approach on eight diverse real-world manipulation tasks. It achieves an average success rate of 96.3% within 45-90 minutes of online fine-tuning, outperforming prior supervised methods with a 144% improvement in success rate and 1.9x shorter episode length. This work highlights the potential of integrating reinforcement learning to enhance the performance of VLA models for real-world robotic applications. Videos and code are available at our project website https://cccedric.github.io/conrft/.

Vision-Language-Action (VLA) models aim to control robots for manipulation from visual observations and natural-language instructions. However, existing hierarchical and autoregressive paradigms often introduce architectural overhead, suffer from temporal inconsistency and long-horizon error accumulation, and lack a mechanism to capture environment dynamics without extra modules. To this end, we present MMaDA-VLA, a fully native pre-trained large diffusion VLA model that unifies multi-modal understanding and generation in a single framework. Our key idea is a native discrete diffusion formulation that embeds language, images, and continuous robot controls into one discrete token space and trains a single backbone with masked token denoising to jointly generate a future goal observation and an action chunk in parallel. Iterative denoising enables global, order-free refinement, improving long-horizon consistency while grounding actions in predicted future visual outcomes without auxiliary world models. Experiments across simulation benchmarks and real-world tasks show state-of-the-art performance, achieving 98.0% average success on LIBERO and 4.78 average length on CALVIN.

Recent large visual-language action models pretrained on diverse robot datasets have demonstrated the potential for generalizing to new environments with a few in-domain data. However, those approaches usually predict discretized or continuous actions by a small action head, which limits the ability in handling diverse action spaces. In contrast, we model the continuous action with a large multi-modal diffusion transformer, dubbed as Diffusion Transformer Policy, in which we directly denoise action chunks by a large transformer model rather than a small action head. By leveraging the scaling capability of transformers, the proposed approach can effectively model continuous end-effector actions across large diverse robot datasets, and achieve better generalization performance. Extensive experiments demonstrate Diffusion Transformer Policy pretrained on diverse robot data can generalize to different embodiments, including simulation environments like Maniskill2 and Calvin, as well as the real-world Franka arm. Specifically, without bells and whistles, the proposed approach achieves state-of-the-art performance with only a single third-view camera stream in the Calvin novel task setting (ABC->D), improving the average number of tasks completed in a row of 5 to 3.6, and the pretraining stage significantly facilitates the success sequence length on the Calvin by over 1.2. The code will be publicly available.

Recent advances in vision-language-action (VLA) models have shown promise in integrating image generation with action prediction to improve generalization and reasoning in robot manipulation. However, existing methods are limited to challenging image-based forecasting, which suffers from redundant information and lacks comprehensive and critical world knowledge, including dynamic, spatial and semantic information. To address these limitations, we propose DreamVLA, a novel VLA framework that integrates comprehensive world knowledge forecasting to enable inverse dynamics modeling, thereby establishing a perception-prediction-action loop for manipulation tasks. Specifically, DreamVLA introduces a dynamic-region-guided world knowledge prediction, integrated with the spatial and semantic cues, which provide compact yet comprehensive representations for action planning. This design aligns with how humans interact with the world by first forming abstract multimodal reasoning chains before acting. To mitigate interference among the dynamic, spatial and semantic information during training, we adopt a block-wise structured attention mechanism that masks their mutual attention, preventing information leakage and keeping each representation clean and disentangled. Moreover, to model the conditional distribution over future actions, we employ a diffusion-based transformer that disentangles action representations from shared latent features. Extensive experiments on both real-world and simulation environments demonstrate that DreamVLA achieves 76.7% success rate on real robot tasks and 4.44 average length on the CALVIN ABC-D benchmarks.

A fundamental requirement for real-world robotic deployment is the ability to understand and respond to natural language instructions. Existing language-conditioned manipulation tasks typically assume that instructions are perfectly aligned with the environment. This assumption limits robustness and generalization in realistic scenarios where instructions may be ambiguous, irrelevant, or infeasible. To address this problem, we introduce RAtional MAnipulation (RAMA), a new benchmark that challenges models with both unseen executable instructions and defective ones that should be rejected. In RAMA, we construct a dataset with over 14,000 samples, including diverse defective instructions spanning six dimensions: visual, physical, semantic, motion, safety, and out-of-context. We further propose the Rational Vision-Language-Action model (RationalVLA). It is a dual system for robotic arms that integrates the high-level vision-language model with the low-level manipulation policy by introducing learnable latent space embeddings. This design enables RationalVLA to reason over instructions, reject infeasible commands, and execute manipulation effectively. Experiments demonstrate that RationalVLA outperforms state-of-the-art baselines on RAMA by a 14.5% higher success rate and 0.94 average task length, while maintaining competitive performance on standard manipulation tasks. Real-world trials further validate its effectiveness and robustness in practical applications. Our project page is https://irpn-eai.github.io/RationalVLA.

Recent advancements in vision-language-action (VLA) models have shown promise in robotic manipulation, yet they continue to struggle with long-horizon, multi-step tasks. Existing methods lack internal reasoning mechanisms that can identify task-relevant interaction cues or track progress within a subtask, leading to critical execution errors such as repeated actions, missed steps, and premature termination. To address these challenges, we introduce PALM, a VLA framework that structures policy learning around interaction-centric affordance reasoning and subtask progress cues. PALM distills complementary affordance representations that capture object relevance, contact geometry, spatial placements, and motion dynamics, and serve as task-relevant anchors for visuomotor control. To further stabilize long-horizon execution, PALM predicts continuous within-subtask progress, enabling seamless subtask transitions. Across extensive simulation and real-world experiments, PALM consistently outperforms baselines, achieving a 91.8% success rate on LIBERO-LONG, a 12.5% improvement in average length on CALVIN ABC->D, and a 2x improvement over real-world baselines across three long-horizon generalization settings.

Large language models have achieved remarkable progress on complex reasoning tasks. However, they often implicitly fabricate information when inputs are incomplete, producing confident but unreliable conclusions -- a failure mode we term ungrounded reasoning. We argue that this issue arises not from insufficient reasoning capability, but from the lack of inferential boundary awareness -- the ability to recognize when the necessary premises for valid inference are missing. To address this issue, we propose Grounded Reasoning via Interactive Reinforcement Learning (GRIL), a multi-turn reinforcement learning framework for grounded reasoning under incomplete information. GRIL decomposes the reasoning process into two stages: clarify and pause, which identifies whether the available information is sufficient, and grounded reasoning, which performs task solving once the necessary premises are established. We design stage-specific rewards to penalize hallucinations, enabling models to detect gaps, stop proactively, and resume reasoning after clarification. Experiments on GSM8K-Insufficient and MetaMATH-Insufficient show that GRIL significantly improves premise detection (up to 45%), leading to a 30% increase in task success while reducing average response length by over 20%. Additional analyses confirm robustness to noisy user responses and generalization to out-of-distribution tasks.

Vision-language-action (VLA) models have shown great potential in building generalist robots, but still face a dilemma-misalignment of 2D image forecasting and 3D action prediction. Besides, such a vision-action entangled training manner limits model learning from large-scale, action-free web video data. To address these issues, we propose DeFI, a novel framework that Decouples visual Forward and Inverse dynamics pretraining to exploit respective data sources, wherein video generation and action prediction are disentangled. We introduce the General Forward Dynamics Model (GFDM), pretrained on diverse human and robot videos for future prediction, and the General Inverse Dynamics Model (GIDM), trained via self-supervised learning to infer latent actions from unlabeled video transitions. These models are then integrated into a unified architecture for end-to-end finetuning on downstream tasks. In this manner, GFDM and GIDM first shine separately and then cooperate for mutual benefit. Extensive experiments on CALVIN ABC-D and SimplerEnv demonstrate state-of-the-art performance, with DeFI achieving an average task length of 4.51 for CALVIN, 51.2% success rate on SimplerEnv-Fractal benchmark and 81.3% success rate in real-world deployment, significantly outperforming prior methods.

Learning generalizable vision-language-action (VLA) models from large-scale human videos is promising but challenging due to cross-embodiment discrepancies in both visual observations and executable actions. While latent action models reduce the action execution gap by learning action abstractions, they still rely on visual features. Thus, misaligned human and robot visual representations can lead to inconsistencies in policy inputs and induce domain-dependent latent actions, hindering effective co-training with human videos. To address this, we propose HARP, a human-robot aligned representation learning framework for more effective VLA pretraining from human videos. Specifically, HARP uses limited paired human-robot demonstrations as cross-embodiment bridges and abundant unpaired human and robot videos as a scalable dynamics supervision data source. It trains a robot-adapted visual encoder and a latent action model with manipulation-centric auxiliary cues and a source-relative pair-discriminative alignment loss, which adapts robot representations toward human semantics while preserving pair-level discrimination. The learned aligned vision encoder and latent action model provide a unified vision and action representation for VLA-style policy learning, where human and robot videos provide vision-language-to-latent-action supervision and a lightweight robot action head grounds latent actions into executable commands. Experiments on feature visualization, simulation, and realworld manipulation show improved human-robot alignment and downstream policy performance, achieving 4.481 average length on CALVIN ABCrightarrowD and a 7.1\% realworld success rate gain over the strongest baseline.

Latent world models predict future states for goal-directed planning using action chunks spanning multiple primitive steps. Existing methods typically use fixed-length chunks and either omit goal-conditioned action generation or limit their supervision to short goal spans. We introduce FlexiWorld, a JEPA-based world model that combines mixed-span goal supervision with variable-length action chunks to improve long-horizon control. During training, we sample varying goal spans and randomly partition the actions into variable-length chunks. We jointly train the world model with a causal action encoder that embeds variable-length chunks and an autoregressive actor that generates primitive actions sequentially. Student Forcing reduces exposure bias by training on generated action prefixes. For planning, Actor-Residual Cross-Entropy Method (ARCEM) combines action-residual search with within-chunk autoregressive feedback and chunk-boundary latent prediction. Across four benchmarks and goal distances, FlexiWorld with ARCEM achieves 89.29% mean success, compared with 83.98% for the strongest baseline. PushT ablations show improved direct control from mixed-span supervision, variable-length chunks, and Student Forcing. Without retraining, FlexiWorld supports different planning chunk lengths: longer chunks accelerate ARCEM by approximately 1.3times on average while maintaining comparable average success.

Great successes have been reported using Reinforcement Learning from Human Feedback (RLHF) to align large language models. Open-source preference datasets and reward models have enabled wider experimentation beyond generic chat settings, particularly to make systems more "helpful" for tasks like web question answering, summarization, and multi-turn dialogue. When optimizing for helpfulness, RLHF has been consistently observed to drive models to produce longer outputs. This paper demonstrates that optimizing for response length is a significant factor behind RLHF's reported improvements in these settings. First, we study the relationship between reward and length for reward models trained on three open-source preference datasets for helpfulness. Here, length correlates strongly with reward, and improvements in reward score are driven in large part by shifting the distribution over output lengths. We then explore interventions during both RL and reward model learning to see if we can achieve the same downstream improvements as RLHF without increasing length. While our interventions mitigate length increases, they aren't uniformly effective across settings. Furthermore, we find that even running RLHF with a reward based solely on length can reproduce most of the downstream improvements over the initial policy model, showing that reward models in these settings have a long way to go.

In this work, we study the issue of reward hacking on the response length, a challenge emerging in Reinforcement Learning from Human Feedback (RLHF) on LLMs. A well-formatted, verbose but less helpful response from the LLMs can often deceive LLMs or even human evaluators to achieve high scores. The same issue also holds for some reward models in RL. To address the challenges in both training and evaluation, we establish a more reliable evaluation protocol for comparing different training configurations, which inspects the trade-off between LLM evaluation score and response length obtained by varying training hyperparameters. Based on this evaluation, we conduct large-scale studies, where the results shed insights into the efficacy of hyperparameters and tricks used in RL on mitigating length bias. We further propose to improve the reward model by jointly training two linear heads on shared feature representations to predict the rewards, one trained to correlate with length, and the other trained to decorrelate with length and therefore focus more on the actual content. We then discard the length head in RL to prevent reward hacking on length. Experiments demonstrate that our approach almost eliminates the reward correlation with length, and improves the obtained policy by a significant margin.

Recent research across mathematical problem solving, proof assistant programming and multimodal jailbreaking documents a striking finding: when (multimodal) language model tackle a suite of tasks with multiple attempts per task -- succeeding if any attempt is correct -- then the negative log of the average success rate scales a power law in the number of attempts. In this work, we identify an apparent puzzle: a simple mathematical calculation predicts that on each problem, the failure rate should fall exponentially with the number of attempts. We confirm this prediction empirically, raising a question: from where does aggregate polynomial scaling emerge? We then answer this question by demonstrating per-problem exponential scaling can be made consistent with aggregate polynomial scaling if the distribution of single-attempt success probabilities is heavy tailed such that a small fraction of tasks with extremely low success probabilities collectively warp the aggregate success trend into a power law - even as each problem scales exponentially on its own. We further demonstrate that this distributional perspective explains previously observed deviations from power law scaling, and provides a simple method for forecasting the power law exponent with an order of magnitude lower relative error, or equivalently, {sim}2-4 orders of magnitude less inference compute. Overall, our work contributes to a better understanding of how neural language model performance improves with scaling inference compute and the development of scaling-predictable evaluations of (multimodal) language models.

The robot learning community has made great strides in recent years, proposing new architectures and showcasing impressive new capabilities; however, the dominant metric used in the literature, especially for physical experiments, is "success rate", i.e. the percentage of runs that were successful. Furthermore, it is common for papers to report this number with little to no information regarding the number of runs, the initial conditions, and the success criteria, little to no narrative description of the behaviors and failures observed, and little to no statistical analysis of the findings. In this paper we argue that to move the field forward, researchers should provide a nuanced evaluation of their methods, especially when evaluating and comparing learned policies on physical robots. To do so, we propose best practices for future evaluations: explicitly reporting the experimental conditions, evaluating several metrics designed to complement success rate, conducting statistical analysis, and adding a qualitative description of failures modes. We illustrate these through an evaluation on physical robots of several learned policies for manipulation tasks.

This paper explores the impact of extending input lengths on the capabilities of Large Language Models (LLMs). Despite LLMs advancements in recent times, their performance consistency across different input lengths is not well understood. We investigate this aspect by introducing a novel QA reasoning framework, specifically designed to assess the impact of input length. We isolate the effect of input length using multiple versions of the same sample, each being extended with padding of different lengths, types and locations. Our findings show a notable degradation in LLMs' reasoning performance at much shorter input lengths than their technical maximum. We show that the degradation trend appears in every version of our dataset, although at different intensities. Additionally, our study reveals that traditional perplexity metrics do not correlate with performance of LLMs' in long input reasoning tasks. We analyse our results and identify failure modes that can serve as useful guides for future research, potentially informing strategies to address the limitations observed in LLMs.

Machine translation tests masked diffusion language models (dLLMs) because every source token must be rendered faithfully, while fixed canvas decoding must choose target length before denoising. Existing masked diffusion decoding work mainly studies token unmasking order, leaving this length decision under-explored despite its direct effect on coverage and redundancy. We introduce Entropy-Valley (EV), a training-free length selector that scores candidate target canvases by mean predictive entropy from all-mask forward passes and selects the canvas the backbone is most prepared to fill. Relative to a baseline using training corpus length statistics, EV recovers 64.9%, 65.3%, and 33.0% of the COMET-22 gain from reference target lengths on EntoZh, ZhtoEn, and EntoDe. Our diagnostics show that denoising-friendly lengths need not match reference lengths. Evaluation by three translation experts supports the EnleftrightarrowZh adequacy gains, with stronger evidence on ZhtoEn. Compared with a LLaMA-3-8B autoregressive (AR) model trained on the same fine-tuning data, the EV system ties on EntoZh and leads on ZhtoEn; an oracle-length diagnostic further shows that, in this masked diffusion MT setting, deciding which tokens to reveal first matters less than how the target length is supplied.

We present explicit formulas - that are also computer code - for 101 real-life quantitative trading alphas. Their average holding period approximately ranges 0.6-6.4 days. The average pair-wise correlation of these alphas is low, 15.9%. The returns are strongly correlated with volatility, but have no significant dependence on turnover, directly confirming an earlier result based on a more indirect empirical analysis. We further find empirically that turnover has poor explanatory power for alpha correlations.

The instruction-following ability of large language models enables humans to interact with AI agents in a natural way. However, when required to generate responses of a specific length, large language models often struggle to meet users' needs due to their inherent difficulty in accurately perceiving numerical constraints. To explore the ability of large language models to control the length of generated responses, we propose the Target Length Generation Task (TLG) and design two metrics, Precise Match (PM) and Flexible Match (FM) to evaluate the model's performance in adhering to specified response lengths. Furthermore, we introduce a novel, model-agnostic approach called Ruler, which employs Meta Length Tokens (MLTs) to enhance the instruction-following ability of large language models under length-constrained instructions. Specifically, Ruler equips LLMs with the ability to generate responses of a specified length based on length constraints within the instructions. Moreover, Ruler can automatically generate appropriate MLT when length constraints are not explicitly provided, demonstrating excellent versatility and generalization. Comprehensive experiments show the effectiveness of Ruler across different LLMs on Target Length Generation Task, e.g., at All Level 27.97 average gain on PM, 29.57 average gain on FM. In addition, we conduct extensive ablation experiments to further substantiate the efficacy and generalization of Ruler. Our code and data is available at https://github.com/Geaming2002/Ruler.

To enable building and testing models on long-document comprehension, we introduce QuALITY, a multiple-choice QA dataset with context passages in English that have an average length of about 5,000 tokens, much longer than typical current models can process. Unlike in prior work with passages, our questions are written and validated by contributors who have read the entire passage, rather than relying on summaries or excerpts. In addition, only half of the questions are answerable by annotators working under tight time constraints, indicating that skimming and simple search are not enough to consistently perform well. Our baseline models perform poorly on this task (55.4%) and significantly lag behind human performance (93.5%).

A unique aspect of ColBERT is its use of [MASK] tokens in queries to score documents (query augmentation). Prior work shows [MASK] tokens weighting non-[MASK] query terms, emphasizing certain tokens over others , rather than introducing whole new terms as initially proposed. We begin by demonstrating that a term weighting behavior previously reported for [MASK] tokens in ColBERTv1 holds for ColBERTv2. We then examine the effect of changing the number of [MASK] tokens from zero to up to four times past the query input length used in training, both for first stage retrieval, and for scoring candidates, observing an initial decrease in performance with few [MASK]s, a large increase when enough [MASK]s are added to pad queries to an average length of 32, then a plateau in performance afterwards. Additionally, we compare baseline performance to performance when the query length is extended to 128 tokens, and find that differences are small (e.g., within 1% on various metrics) and generally statistically insignificant, indicating performance does not collapse if ColBERT is presented with more [MASK] tokens than expected.

Recently, recurrent models such as state space models and linear attention have become popular due to their linear complexity in the sequence length. Thanks to their recurrent nature, in principle they can process arbitrarily long sequences, but their performance sometimes drops considerably beyond their training context lengths-i.e. they fail to length generalize. In this work, we provide comprehensive empirical and theoretical analysis to support the unexplored states hypothesis, which posits that models fail to length generalize when during training they are only exposed to a limited subset of the distribution of all attainable states (i.e. states that would be attained if the recurrence was applied to long sequences). Furthermore, we investigate simple training interventions that aim to increase the coverage of the states that the model is trained on, e.g. by initializing the state with Gaussian noise or with the final state of a different input sequence. With only 500 post-training steps (sim 0.1% of the pre-training budget), these interventions enable length generalization for sequences that are orders of magnitude longer than the training context (e.g. 2klongrightarrow 128k) and show improved performance in long context tasks, thus presenting a simple and efficient way to enable robust length generalization in general recurrent models.

Large language models (LLMs) are increasingly optimized for long reasoning, under the assumption that more reasoning leads to better performance. However, emerging evidence suggests that longer responses can sometimes degrade accuracy rather than improve it. In this paper, we conduct a systematic empirical study of the relationship between reasoning length and answer correctness. We find that LLMs tend to overthink simple problems, generating unnecessarily long outputs, and underthink harder ones, failing to extend their reasoning when it is most needed. This indicates that models might misjudge problem difficulty and fail to calibrate their response length appropriately. Furthermore, we investigate the effects of length reduction with a preference optimization algorithm when simply preferring the shorter responses regardless of answer correctness. Experiments show that the generation length can be significantly reduced while maintaining acceptable accuracy. Our findings highlight generation length as a meaningful signal for reasoning behavior and motivate further exploration into LLMs' self-awareness in reasoning length adaptation.

We study the relationship between vocabulary size and text length in a corpus of 75 literary works in English, authored by six writers, distinguishing between the contributions of three grammatical classes (or ``tags,'' namely, {\it nouns}, {\it verbs}, and {\it others}), and analyze the progressive appearance of new words of each tag along each individual text. While the power-law relation prescribed by Heaps' law is satisfactorily fulfilled by total vocabulary sizes and text lengths, the appearance of new words in each text is on the whole well described by the average of random shufflings of the text, which does not obey a power law. Deviations from this average, however, are statistically significant and show a systematic trend across the corpus. Specifically, they reveal that the appearance of new words along each text is predominantly retarded with respect to the average of random shufflings. Moreover, different tags are shown to add systematically distinct contributions to this tendency, with {\it verbs} and {\it others} being respectively more and less retarded than the mean trend, and {\it nouns} following instead this overall mean. These statistical systematicities are likely to point to the existence of linguistically relevant information stored in the different variants of Heaps' law, a feature that is still in need of extensive assessment.

Recent popularity surrounds large AI language models due to their impressive natural language capabilities. They contribute significantly to language-related tasks, including prompt-based learning, making them valuable for various specific tasks. This approach unlocks their full potential, enhancing precision and generalization. Research communities are actively exploring their applications, with ChatGPT receiving recognition. Despite extensive research on large language models, their potential in recommendation scenarios still needs to be explored. This study aims to fill this gap by investigating ChatGPT's capabilities as a zero-shot recommender system. Our goals include evaluating its ability to use user preferences for recommendations, reordering existing recommendation lists, leveraging information from similar users, and handling cold-start situations. We assess ChatGPT's performance through comprehensive experiments using three datasets (MovieLens Small, Last.FM, and Facebook Book). We compare ChatGPT's performance against standard recommendation algorithms and other large language models, such as GPT-3.5 and PaLM-2. To measure recommendation effectiveness, we employ widely-used evaluation metrics like Mean Average Precision (MAP), Recall, Precision, F1, normalized Discounted Cumulative Gain (nDCG), Item Coverage, Expected Popularity Complement (EPC), Average Coverage of Long Tail (ACLT), Average Recommendation Popularity (ARP), and Popularity-based Ranking-based Equal Opportunity (PopREO). Through thoroughly exploring ChatGPT's abilities in recommender systems, our study aims to contribute to the growing body of research on the versatility and potential applications of large language models. Our experiment code is available on the GitHub repository: https://github.com/sisinflab/Recommender-ChatGPT

An agent that inherits six one-line memories may pull at most one archived source record before acting; a directive written into the store can steer that choice: a pointer to the record, a criterion that identifies it, or both. Across twelve registered studies on one instrument lineage (14,760 attempts) we measured where the request goes under each form. On six direct-provider models a length-matched criterion exceeded a bare id by +35.0 points [+31.2, +38.8] (Study D); the contrast failed its registered superiority rule on a nine-model OpenRouter-served panel (Study E). Appending the id cancelled the criterion on three Claude models (Opus 5: 40/40 to 0/40; Study F-x); six byte-matched edits gave each exact string its own effect (Study G), and a re-run at eighty runs per cell left fifteen of thirty replication contrasts within the margin, fifteen unresolved and none beyond (Study G'). A ratification line (+96.0 points on Opus 5) and a budget of two credits restored the target on all three (Study J); across five criterion strings the suffix's cancellation held for four of the five wordings on Opus 5 and all five wordings on Fable 5.1 (Study H2); in a second store every model followed the criterion (Study H1). Continued into a decision, the criterion moved the choice toward the current record (+100.0 points, Opus 5) and away from it on Fable 5.1 (Study I). A one-character plan pointer's effect (+78.0 points; Study B, after a correction of its first repository report) returned the same verdict under a prospectively registered re-run (+81.7 points; Study B'). All results are descriptive effects of exact edits on fixed panels with registered intervals and no mechanism claim.

Many applications involve estimating the mean of multiple binomial outcomes as a common problem -- assessing intergenerational mobility of census tracts, estimating prevalence of infectious diseases across countries, and measuring click-through rates for different demographic groups. The most standard approach is to report the plain average of each outcome. Despite simplicity, the estimates are noisy when the sample sizes or mean parameters are small. In contrast, the Empirical Bayes (EB) methods are able to boost the average accuracy by borrowing information across tasks. Nevertheless, the EB methods require a Bayesian model where the parameters are sampled from a prior distribution which, unlike the commonly-studied Gaussian case, is unidentified due to discreteness of binomial measurements. Even if the prior distribution is known, the computation is difficult when the sample sizes are heterogeneous as there is no simple joint conjugate prior for the sample size and mean parameter. In this paper, we consider the compound decision framework which treats the sample size and mean parameters as fixed quantities. We develop an approximate Stein's Unbiased Risk Estimator (SURE) for the average mean squared error given any class of estimators. For a class of machine learning-assisted linear shrinkage estimators, we establish asymptotic optimality, regret bounds, and valid inference. Unlike existing work, we work with the binomials directly without resorting to Gaussian approximations. This allows us to work with small sample sizes and/or mean parameters in both one-sample and two-sample settings. We demonstrate our approach using three datasets on firm discrimination, education outcomes, and innovation rates.

Scaling law principles indicate a power-law correlation between loss and variables such as model size, dataset size, and computational resources utilized during training. These principles play a vital role in optimizing various aspects of model pre-training, ultimately contributing to the success of large language models such as GPT-4, Llama and Gemini. However, the original scaling law paper by OpenAI did not disclose the complete details necessary to derive the precise scaling law formulas, and their conclusions are only based on models containing up to 1.5 billion parameters. Though some subsequent works attempt to unveil these details and scale to larger models, they often neglect the training dependency of important factors such as the learning rate, context length and batch size, leading to their failure to establish a reliable formula for predicting the test loss trajectory. In this technical report, we confirm that the scaling law formulations proposed in the original OpenAI paper remain valid when scaling the model size up to 33 billion, but the constant coefficients in these formulas vary significantly with the experiment setup. We meticulously identify influential factors and provide transparent, step-by-step instructions to estimate all constant terms in scaling-law formulas by training on models with only 1M~60M parameters. Using these estimated formulas, we showcase the capability to accurately predict various attributes for models with up to 33B parameters before their training, including (1) the minimum possible test loss; (2) the minimum required training steps and processed tokens to achieve a specific loss; (3) the critical batch size with an optimal time/computation trade-off at any loss value; and (4) the complete test loss trajectory with arbitrary batch size.

Detailed profession-specific system prompts raise token use and estimated cost per response without a consistent accuracy gain. We evaluate Scientific Agents, an open-source corpus of 503 profession-specific AGENTS.md profiles, with Gemini 3.8 Flash via OpenRouter in the Pi agent harness. We compare matched profiles with four controls: a minimal baseline ("You are a helpful assistant"), the profile's opening role sentence, a generic scientific rigor guide, and a profile from an unrelated domain. Across nine text-based science benchmarks (4,531 sampled questions, 100 matched profiles), 4,488 items completed all five conditions after API-error retries, scored with automated, rule-based grading. The average profile-baseline accuracy difference is -0.6 percentage points (95% bootstrap interval [-1.5, +0.2] across fixed tasks), and no benchmark shows a statistically clear improvement. Matched profiles produced 1.5-2.3 times as many output tokens and cost 2.2-4.5 times more per successful call. On 60 tool-using BioMysteryBench bioinformatics problems (three runs each for baseline and profile), mean solve rates were 46.7% with the profile and 56.7% at baseline, a difference of -10.0 percentage points (95% interval [-16.7, -3.3]) driven by more frequent token- and time-limit stops under the profile. Longer prompts had one unexpected operational advantage: on SuperGPQA, frequent provider API drops left the short baseline with a correct first-pass answer on only 54.0% of items, against 71.6% with the profile. Generic and mismatched prompts were about as reliable, so this gain comes from prompt length or formatting rather than domain expertise. For the tested model and tasks, loading full profession profiles by default does not improve accuracy and costs considerably more; whether selective retrieval of profile sections or open-ended scientific tasks would change this remains to be tested.

One straightforward metric to evaluate a survival prediction model is based on the Mean Absolute Error (MAE) -- the average of the absolute difference between the time predicted by the model and the true event time, over all subjects. Unfortunately, this is challenging because, in practice, the test set includes (right) censored individuals, meaning we do not know when a censored individual actually experienced the event. In this paper, we explore various metrics to estimate MAE for survival datasets that include (many) censored individuals. Moreover, we introduce a novel and effective approach for generating realistic semi-synthetic survival datasets to facilitate the evaluation of metrics. Our findings, based on the analysis of the semi-synthetic datasets, reveal that our proposed metric (MAE using pseudo-observations) is able to rank models accurately based on their performance, and often closely matches the true MAE -- in particular, is better than several alternative methods.

The abilities of long-context language models (LMs) are often evaluated using the "Needle-in-a-Haystack" (NIAH) test, which comprises tasks designed to assess a model's ability to identify specific information ("needle") within large text sequences ("haystack"). While these benchmarks measure how well models understand long-context input sequences, they do not effectively gauge the quality of long-form text generation--a critical aspect for applications such as design proposals and creative writing. To address this gap, we have introduced a new long-form text evaluation benchmark, Spinning the Golden Thread (SGT), which tests models' ability to identify specific events within generated long text sequences. In this benchmark, we prompt long-context LMs to create long-form text that must include particular events or constraints and evaluate their ability to incorporate these elements. We evaluated ten long-context LMs across four distinct scenarios, three types of prompt instructions, and two different generation-length settings (16K and 32K). Although these models perform well on NIAH benchmarks, none demonstrated satisfactory performance on the Spinning the Golden Thread, raising concerns about their ability to generate coherent long-form text that follows instructions. Additionally, as the length of the generated text increases, all models exhibit a significant drop in performance.

This study investigates retrieval-augmented summarization by specifically examining the impact of exemplar summary lengths under length constraints, not covered by previous work. We propose a Diverse Length-aware Maximal Marginal Relevance (DL-MMR) algorithm to better control summary lengths. This algorithm combines the query relevance with diverse target lengths in retrieval-augmented summarization. Unlike previous methods that necessitate exhaustive exemplar exemplar relevance comparisons using MMR, DL-MMR considers the exemplar target length as well and avoids comparing exemplars to each other, thereby reducing computational cost and conserving memory during the construction of an exemplar pool. Experimental results showed the effectiveness of DL-MMR, which considers length diversity, compared to the original MMR algorithm. DL-MMR additionally showed the effectiveness in memory saving of 781,513 times and computational cost reduction of 500,092 times, while maintaining the same level of informativeness.

Large Language Models have achieved remarkable success in natural language processing tasks, with Reinforcement Learning playing a key role in adapting them to specific applications. However, obtaining ground truth answers for training LLMs in mathematical problem-solving is often challenging, costly, and sometimes unfeasible. This research delves into the utilization of format and length as surrogate signals to train LLMs for mathematical problem-solving, bypassing the need for traditional ground truth answers.Our study shows that a reward function centered on format correctness alone can yield performance improvements comparable to the standard GRPO algorithm in early phases. Recognizing the limitations of format-only rewards in the later phases, we incorporate length-based rewards. The resulting GRPO approach, leveraging format-length surrogate signals, not only matches but surpasses the performance of the standard GRPO algorithm relying on ground truth answers in certain scenarios, achieving 40.0\% accuracy on AIME2024 with a 7B base model. Through systematic exploration and experimentation, this research not only offers a practical solution for training LLMs to solve mathematical problems and reducing the dependence on extensive ground truth data collection, but also reveals the essence of why our label-free approach succeeds: base model is like an excellent student who has already mastered mathematical and logical reasoning skills, but performs poorly on the test paper, it simply needs to develop good answering habits to achieve outstanding results in exams , in other words, to unlock the capabilities it already possesses.

In this paper, we introduce a novel predict-and-optimize method for profit-driven churn prevention. We frame the task of targeting customers for a retention campaign as a regret minimization problem. The main objective is to leverage individual customer lifetime values (CLVs) to ensure that only the most valuable customers are targeted. In contrast, many profit-driven strategies focus on churn probabilities while considering average CLVs. This often results in significant information loss due to data aggregation. Our proposed model aligns with the guidelines of Predict-and-Optimize (PnO) frameworks and can be efficiently solved using stochastic gradient descent methods. Results from 12 churn prediction datasets underscore the effectiveness of our approach, which achieves the best average performance compared to other well-established strategies in terms of average profit.

Recent large language models (LLMs) support long contexts ranging from 128K to 1M tokens. A popular method for evaluating these capabilities is the needle-in-a-haystack (NIAH) test, which involves retrieving a "needle" (relevant information) from a "haystack" (long irrelevant context). Extensions of this approach include increasing distractors, fact chaining, and in-context reasoning. However, in these benchmarks, models can exploit existing literal matches between the needle and haystack to simplify the task. To address this, we introduce NoLiMa, a benchmark extending NIAH with a carefully designed needle set, where questions and needles have minimal lexical overlap, requiring models to infer latent associations to locate the needle within the haystack. We evaluate 12 popular LLMs that claim to support contexts of at least 128K tokens. While they perform well in short contexts (<1K), performance degrades significantly as context length increases. At 32K, for instance, 10 models drop below 50% of their strong short-length baselines. Even GPT-4o, one of the top-performing exceptions, experiences a reduction from an almost-perfect baseline of 99.3% to 69.7%. Our analysis suggests these declines stem from the increased difficulty the attention mechanism faces in longer contexts when literal matches are absent, making it harder to retrieve relevant information.

Large Language Models (LLMs) such as ChatGPT, Claude, and Gemini increasingly act as general-purpose copilots, yet they often respond with unnecessary length on simple requests, adding redundant explanations, hedging, or boilerplate that increases cognitive load and inflates token-based inference cost. Prior work suggests that preference-based post-training and LLM-judged evaluations can induce systematic length bias, where longer answers are rewarded even at comparable quality. We introduce YapBench, a lightweight benchmark for quantifying user-visible over-generation on brevity-ideal prompts. Each item consists of a single-turn prompt, a curated minimal-sufficient baseline answer, and a category label. Our primary metric, YapScore, measures excess response length beyond the baseline in characters, enabling comparisons across models without relying on any specific tokenizer. We summarize model performance via the YapIndex, a uniformly weighted average of category-level median YapScores. YapBench contains over three hundred English prompts spanning three common brevity-ideal settings: (A) minimal or ambiguous inputs where the ideal behavior is a short clarification, (B) closed-form factual questions with short stable answers, and (C) one-line coding tasks where a single command or snippet suffices. Evaluating 76 assistant LLMs, we observe an order-of-magnitude spread in median excess length and distinct category-specific failure modes, including vacuum-filling on ambiguous inputs and explanation or formatting overhead on one-line technical requests. We release the benchmark and maintain a live leaderboard for tracking verbosity behavior over time.

Collecting high-quality training examples for language model fine-tuning is expensive, with practical budgets limiting the amount of data that can be procured. We investigate whether example difficulty affects GRPO training effectiveness by comparing selection strategies (easy, medium, hard, random) across multiple models and reasoning tasks. Training on the hardest 10\% of examples (those where the base model fails most often) yields dramatic performance gains up to 47\%, while easy examples produce minimal improvements of 3-15\%. This occurs because GRPO requires outcome variance to generate learning signals; hard examples maintain mixed success/failure outcomes throughout training while easy examples quickly converge to consistent success, eliminating learning opportunities. Moreover, models trained on hard examples show superior out-of-distribution generalization, with only hard-trained models achieving meaningful gains on the AIME2025 benchmark. Our findings provide clear guidance: when budget-constrained, prioritize collecting and annotating examples where your base model struggles, as these drive nearly all learning value in GRPO fine-tuning

The success of Deepseek-R1 has drawn the LLM community's attention to reinforcement learning (RL) methods like GRPO. However, such rule-based 0/1 outcome reward methods lack the capability to regulate the intermediate reasoning processes during chain-of-thought (CoT) generation, leading to severe overthinking phenomena. In response, recent studies have designed reward functions to reinforce models' behaviors in producing shorter yet correct completions. Nevertheless, we observe that these length-penalty reward functions exacerbate RL training instability: as the completion length decreases, model accuracy abruptly collapses, often occurring early in training. To address this issue, we propose a simple yet effective solution GRPO-lambda, an efficient and stabilized variant of GRPO, which dynamically adjusts the reward strategy by monitoring the correctness ratio among completions within each query-sampled group. A low correctness ratio indicates the need to avoid length penalty that compromises CoT quality, triggering a switch to length-agnostic 0/1 rewards that prioritize reasoning capability. A high ratio maintains length penalties to boost efficiency. Experimental results show that our approach avoids training instability caused by length penalty while maintaining the optimal accuracy-efficiency trade-off. On the GSM8K, GPQA, MATH-500, AMC 2023, and AIME 2024 benchmarks, it improves average accuracy by 1.48% while reducing CoT sequence length by 47.3%.

Existing benchmarks measure capability -- whether a model succeeds on a single attempt -- but production deployments require reliability -- consistent success across repeated attempts on tasks of varying duration. We show these properties diverge systematically as task duration grows, and that pass@1 on short tasks is structurally blind to this divergence. We introduce a reliability science framework for long-horizon LLM agents with four metrics: Reliability Decay Curve (RDC), Variance Amplification Factor (VAF), Graceful Degradation Score (GDS), and Meltdown Onset Point (MOP). We evaluate 10 models across 23,392 episodes on a 396-task benchmark spanning four duration buckets and three domains. Key findings: (1) reliability decay is domain-stratified -- SE GDS drops from 0.90 to 0.44 while document processing is nearly flat (0.74 to 0.71); (2) VAF bifurcates by capability tier -- high VAF is a capability signature, not an instability signal; (3) capability and reliability rankings diverge substantially, with multi-rank inversions at long horizons; (4) frontier models have the highest meltdown rates (up to 19%) because they attempt ambitious multi-step strategies that sometimes spiral; and (5) memory scaffolds universally hurt long-horizon performance across all 10 models. These results motivate reliability as a first-class evaluation dimension alongside capability.

To schedule LLM inference, the shortest job first (SJF) principle is favorable by prioritizing requests with short output lengths to avoid head-of-line (HOL) blocking. Existing methods usually predict a single output length for each request to facilitate scheduling. We argue that such a point estimate does not match the stochastic decoding process of LLM inference, where output length is uncertain by nature and determined by when the end-of-sequence (EOS) token is sampled. Hence, the output length of each request should be fitted with a distribution rather than a single value. With an in-depth analysis of empirical data and the stochastic decoding process, we observe that output length follows a heavy-tailed distribution and can be fitted with the log-t distribution. On this basis, we propose a simple metric called Tail Inflated Expectation (TIE) to replace the output length in SJF scheduling, which adjusts the expectation of a log-t distribution with its tail probabilities to account for the risk that a request generates long outputs. To evaluate our TIE scheduler, we compare it with three strong baselines, and the results show that TIE reduces the per-token latency by 2.31times for online inference and improves throughput by 1.42times for offline data generation.

Dexterous manipulation policies learned by imitation are typically evaluated for robustness to variation in scenes, objects, or instructions, but their performance across task execution speeds is less often examined. This leaves open how much temporal robustness a learner retains relative to the expert it imitates. We compare an expert and learner under the same task conditions, initial-condition draws, and speedup factors. We instantiate the evaluation in ParcelStow, a contact-rich task in which the robot acquires, reorients, and inserts a parcel. The demonstrations span the speedup range for the manipulation phases after parcel acquisition. A scripted expert and an Action Chunking with Transformers (ACT) policy trained from the expert's demonstrations both achieve 100 percent task success at nominal speed. Their success rates diverge within the demonstrated range: at its maximum, expert success is 84 percent and ACT success is 53 percent. Two ACT policies with different parameter initializations show similar degradation, decreasing by 34 and 48 percentage points from nominal speed to the maximum demonstrated speed, compared with 16 points for the expert. Stage-level analysis shows that 35 of ACT's 47 failures at the maximum demonstrated speed are insertion misalignments. Under the relative-motion handoff, every ACT acquisition retains the parcel through reorientation and transfer in free space, but only 64 percent complete the overall task, compared with 95 percent after expert acquisition. Across all evaluated policies and speeds, none of the 414 acquisitions without force closure completes the task. Equal nominal task success therefore does not imply preservation of expert performance across execution speeds. Code, data, and evaluation scripts are available at https://github.com/coenwerem/parcelstow.

Reinforcement learning substantially improves reasoning in large language models, but it also tends to lengthen chain of thought outputs and increase computational cost during both training and inference. Though length control methods have been proposed, it remains unclear what the optimal output length is for balancing efficiency and performance. In this work, we compare several length control methods on two models, Qwen3-1.7B Base and DeepSeek-R1-Distill-Qwen-1.5B. Our results indicate that length penalties may hinder reasoning acquisition, while properly tuned length control can improve efficiency for models with strong prior reasoning. By extending prior work to RL trained policies, we identify two failure modes, 1) long outputs increase dispersion, and 2) short outputs lead to under-thinking.

This article presents an experiment in fine-tuning a pretrained causal language model (Meta's Llama 3.1 8B Instruct) for aiding in three fundamental tasks of philological research: chronological and geographic attribution as well as text restoration in ancient Greek inscriptions and documentary papyri. Using a prompt-based instruct approach, the fine-tuned models surpass the state of the art in key metrics. For inscriptions, the models achieve a lower average character error rate (CER) of 22.5% (vs. 26.3%), while closely matching top-1 accuracy (60.9% vs. 61.8%) and top-20 accuracy (77.5% vs. 78.3%) for sequences up to 10 characters. They also provide a practical advantage by ignoring spaces during reconstruction, aligning better with the scriptio continua typically used in ancient written artifacts. In geographic attribution, the model outperforms previous benchmarks with a top-1 accuracy of 75.0% (vs. 70.8%) and a top-3 accuracy of 83.7% (vs. 82.1%). For dating, it achieves an average deviation of 26.2 years (vs. 29.3) and a median deviation of 1 year (vs. 3) from the actual date range. The models also set new baselines for documentary papyri, with a CER of 16.3%, a top-1 accuracy of 71.3%, and top-20 of 85.0% in text reconstruction; a top-1 accuracy of 66.4% and top-3 of 79.9% in geographic attribution; and, in chronological attribution, a deviation of 21.7 years from the actual termini post/ante quem, with a median deviation of 0 years.

Efficiently managing compute resources for Large Language Model (LLM) inference remains challenging due to the inherently stochastic and variable lengths of autoregressive text generation. Accurately estimating response lengths in advance enables proactive resource allocation, yet existing approaches either bias text generation towards certain lengths or rely on assumptions that ignore model- and prompt-specific variability. We introduce CASTILLO, a dataset characterizing response length distributions across 13 widely-used open-source LLMs evaluated on seven distinct instruction-following corpora. For each langleprompt, modelrangle sample pair, we generate 10 independent completions using fixed decoding hyper-parameters, record the token length of each response, and publish summary statistics (mean, std-dev, percentiles), along with the shortest and longest completions, and the exact generation settings. Our analysis reveals significant inter- and intra-model variability in response lengths (even under identical generation settings), as well as model-specific behaviors and occurrences of partial text degeneration in only subsets of responses. CASTILLO enables the development of predictive models for proactive scheduling and provides a systematic framework for analyzing model-specific generation behaviors. We publicly release the dataset and code to foster research at the intersection of generative language modeling and systems.

Recursive self-improvement (RSI) asks whether an AI system can improve the process that produces AI systems, so that the next system inherits the improvement. That process is the training algorithm: a better objective or update rule improves the compute-capability exchange rate for every subsequent run, including the one that produces the next agent. Whether RSI is feasible therefore turns on whether an agent can design training algorithms. No benchmark isolates that ability: existing suites are won by collecting data or by tuning hyperparameters, and none tells a change to how a run is executed apart from a change to how the model learns. We present AI4AI-Bench, 10 frozen research repositories spanning 10 training algorithm families. In each task, an agent has 4 hours on one B300 to rewrite the training algorithm; its code is then rerun from scratch for up to 12 hours and scored by a fixed evaluator hidden from the agent, against the repository's original algorithm under the same procedure. Because the 10 metrics are incommensurable, every task is mapped onto one scale on which 0 is an uninformative model, 0.1 is the algorithm the repository ships, and 1.0 is the task optimum. Across 29 configurations of 6 systems on all 10 tasks the mean score is 0.166, and the best system reaches 0.250: even the strongest closes under a fifth of the distance between the algorithm that was already there and the optimum. The submissions show where that distance went: most never change how the model learns at all, and the minority that do average 0.226 against 0.126 for the rest. More reasoning effort mostly buys the willingness to go there, taking that minority from 8% of submissions to 64% and the mean score from 0.094 to 0.196. We release the task suite, the evaluators and every scored submission, so that the measurement can be repeated as these systems change.

LLM-based auto-annotators have become a key component of the LLM development process due to their cost-effectiveness and scalability compared to human-based evaluation. However, these auto-annotators can introduce complex biases that are hard to remove. Even simple, known confounders such as preference for longer outputs remain in existing automated evaluation metrics. We propose a simple regression analysis approach for controlling biases in auto-evaluations. As a real case study, we focus on reducing the length bias of AlpacaEval, a fast and affordable benchmark for chat LLMs that uses LLMs to estimate response quality. Despite being highly correlated with human preferences, AlpacaEval is known to favor models that generate longer outputs. We introduce a length-controlled AlpacaEval that aims to answer the counterfactual question: "What would the preference be if the model's and baseline's output had the same length?". To achieve this, we first fit a generalized linear model to predict the biased output of interest (auto-annotator preferences) based on the mediators we want to control for (length difference) and other relevant features. We then obtain length-controlled preferences by predicting preferences while conditioning the GLM with a zero difference in lengths. Length-controlling not only improves the robustness of the metric to manipulations in model verbosity, we also find that it increases the Spearman correlation with LMSYS' Chatbot Arena from 0.94 to 0.98. We release the code and leaderboard at https://tatsu-lab.github.io/alpaca_eval/ .

Token serves as the fundamental unit of computation in modern autoregressive models, and generation length directly influences both inference cost and reasoning performance. Despite its importance, existing approaches lack fine-grained length modeling, operating primarily at the coarse-grained sequence level. We introduce the Length Value Model (LenVM), a token-level framework that models the remaining generation length. By formulating length modeling as a value estimation problem and assigning a constant negative reward to each generated token, LenVM predicts a bounded, discounted return that serves as a monotone proxy for the remaining generation horizon. This formulation yields supervision that is annotation-free, dense, unbiased, and scalable. Experiments on LLMs and VLMs demonstrate LenVM provides a highly effective signal at inference time. On the LIFEBench exact length matching task, applying LenVM to a 7B model improves the length score from 30.9 to 64.8, significantly outperforming frontier closed-source models. Furthermore, LenVM enables continuous control over the trade off between performance and efficiency. On GSM8K at a budget of 200 tokens, LenVM maintains 63% accuracy compared to 6 percent for token budget baseline. It also accurately predicts total generation length from the prompt boundary. Finally, LenVM's token-level values offer an interpretable view of generation dynamics, revealing how specific tokens shift reasoning toward shorter or longer regimes. Results demonstrate that LenVM supports a broad range of applications and token length can be effectively modeled as a token-level value signal, highlighting the potential of LenVM as a general framework for length modeling and as a length-specific value signal that could support future RL training. Code is available at https://github.com/eric-ai-lab/Length-Value-Model.

This study explores the relationship between textual features and Information Engagement (IE) on digital platforms. It highlights the impact of computational linguistics and analytics on user interaction. The READ model is introduced to quantify key predictors like representativeness, ease of use, affect, and distribution, which forecast engagement levels. The model's effectiveness is validated through AB testing and randomized trials, showing strong predictive performance in participation (accuracy: 0.94), perception (accuracy: 0.85), perseverance (accuracy: 0.81), and overall IE (accuracy: 0.97). While participation metrics are strong, perception and perseverance show slightly lower recall and F1-scores, indicating some challenges. The study demonstrates that modifying text based on the READ model's insights leads to significant improvements. For example, increasing representativeness and positive affect boosts selection rates by 11 percent, raises evaluation averages from 3.98 to 4.46, and improves retention rates by 11 percent. These findings highlight the importance of linguistic factors in IE, providing a framework for enhancing digital text engagement. The research offers practical strategies applicable to fields like education, health, and media.

We introduce WildBench, an automated evaluation framework designed to benchmark large language models (LLMs) using challenging, real-world user queries. WildBench consists of 1,024 tasks carefully selected from over one million human-chatbot conversation logs. For automated evaluation with WildBench, we have developed two metrics, WB-Reward and WB-Score, which are computable using advanced LLMs such as GPT-4-turbo. WildBench evaluation uses task-specific checklists to evaluate model outputs systematically and provides structured explanations that justify the scores and comparisons, resulting in more reliable and interpretable automatic judgments. WB-Reward employs fine-grained pairwise comparisons between model responses, generating five potential outcomes: much better, slightly better, slightly worse, much worse, or a tie. Unlike previous evaluations that employed a single baseline model, we selected three baseline models at varying performance levels to ensure a comprehensive pairwise evaluation. Additionally, we propose a simple method to mitigate length bias, by converting outcomes of ``slightly better/worse'' to ``tie'' if the winner response exceeds the loser one by more than K characters. WB-Score evaluates the quality of model outputs individually, making it a fast and cost-efficient evaluation metric. WildBench results demonstrate a strong correlation with the human-voted Elo ratings from Chatbot Arena on hard tasks. Specifically, WB-Reward achieves a Pearson correlation of 0.98 with top-ranking models. Additionally, WB-Score reaches 0.95, surpassing both ArenaHard's 0.91 and AlpacaEval2.0's 0.89 for length-controlled win rates, as well as the 0.87 for regular win rates.

Self-improving LLM systems propose changes to themselves and keep those that score better on a small evaluation set. We treat this keep-if-better step as selection under measurement noise, model the correlated errors of the candidates in a single decision, and study empirically what happens when the evaluation set is reused. In runs where Qwen models rewrite their own instructions and every candidate is also scored on 600 held-out items, most proposals after the first are harmful, and the model gives the size of the winner's curse of a generation's best candidate. With a prior from a separate pilot, it matches the average overstatement of first-generation commits in native loops, though not setting by setting. In a pre-registered study, the final selection-set score of greedy loops exceeded held-out accuracy by 13 to 20 points with 16 selection items and by 1 to 5 points with 256. Held-out gains grew with the selection set on TREC but not on GSM8K, and the tested acceptance rules did not beat greedy acceptance over whole runs. Gains measured on the selection set also exceeded held-out gains when a current model refined a competent instruction, and in the validation scores of GEPA and MIPROv2. Scoring the starting and the current instruction on 64 items never used for selection removes the average bias of a loop's reported gain, but single estimates remain off by about 6 points. Self-improvement studies should report held-out gains with their uncertainty.

This study presents a hybrid framework for predicting movie success. The framework integrates multi-task learning (MTL), GPT-based sentiment analysis, and Susceptible-Infected-Recovered (SIR) propagation modeling. The study examines limitations in existing approaches. It models static production attributes, information dissemination, and audience sentiment at the same time. The framework uses 5,840 films from 2004 to 2024 and approximate 300,000 user reviews. It shows predictive performance with classification accuracy of 0.964 and regression metrics of MAE 0.388. Ablation analysis indicates component interactions. Selective feature combinations perform better than the comprehensive model. This result questions assumptions about feature integration. The model shows virality patterns between successful and unsuccessful films. Innovations include epidemiological modeling for information diffusion, multidimensional sentiment features from GPT-based analysis, and a shared representation architecture that optimizes multiple success metrics. The framework provides applications in the film production lifecycle. It also contributes to understanding how audience engagement leads to commercial outcomes.

This paper presents an eight-week observational comparison of 68 single-file HTML generations collected across 17 public experiments in the "HTML AI Battle" project between December 10, 2025 and February 4, 2026. Four reasoning model families, GPT, Gemini, Grok, and Claude, were compared under a fixed public-interface protocol with no custom instructions, no personality tuning, and no repair prompts. Each output was evaluated from a rendered browser video using human scores and a Gemini LLM-as-a-judge layer for prompt adherence, functional correctness, and UI quality, then packaged into a standardized social-media protocol spanning X (Twitter), TikTok, and YouTube. The tracker was also used for two supervised predictive analyses: an experiment-level model for 24-hour X impressions and a generation-level model for HTML verbosity. Under this protocol, Claude was the strongest and most consistent family, leading mean performance and winning 9/17 prompts under the primary human weighted score. Longer measured reasoning time was not associated with higher quality overall. Gemini as a judge was significantly more lenient than the human evaluator on functional correctness and overall performance, while stable self-favoring bias remained unresolved. The exploratory X-impressions model remained weak under post-screen cross-validation (MAE = 46,874, R^2 = -0.377), whereas the HTML-lines model performed better, with a model-family-only baseline outperforming prompt-aware alternatives (MAE = 135.2, R^2 = 0.576). Overall, selected pre-publication technical/audio variables were not sufficient to predict 24-hour X reach, while code verbosity was driven much more by model family than by prompt wording. The comparisons remain observational and are limited by public-interface drift, access-path differences, and one primary human scorer.

Analyzing outcomes in long-term cancer survivor studies can be complex. The effects of predictors on the failure process may be difficult to assess over longer periods of time, as the commonly used assumption of proportionality of hazards holding over an extended period is often questionable. In this manuscript, we compare seven different survival models that estimate the hazard rate and the effects of proportional and non-proportional covariates. In particular, we focus on an extension of the the multi-resolution hazard (MRH) estimator, combining a non-proportional hierarchical MRH approach with a data-driven pruning algorithm that allows for computational efficiency and produces robust estimates even in times of few observed failures. Using data from a large-scale randomized prostate cancer clinical trial, we examine patterns of biochemical failure and estimate the time-varying effects of androgen deprivation therapy treatment and other covariates. We compare the impact of different modeling strategies and smoothness assumptions on the estimated treatment effect. Our results show that the benefits of treatment diminish over time, possibly with implications for future treatment protocols.

We extend the context length of Llama-3-8B-Instruct from 8K to 80K via QLoRA fine-tuning. The entire training cycle is super efficient, which takes 8 hours on one 8xA800 (80G) GPU machine. The resulted model exhibits superior performances across a broad range of evaluation tasks, such as NIHS, topic retrieval, and long-context language understanding; meanwhile, it also well preserves the original capability over short contexts. The dramatic context extension is mainly attributed to merely 3.5K synthetic training samples generated by GPT-4 , which indicates the LLMs' inherent (yet largely underestimated) potential to extend its original context length. In fact, the context length could be extended far beyond 80K with more computation resources. Therefore, the team will publicly release the entire resources (including data, model, data generation pipeline, training code) so as to facilitate the future research from the community: https://github.com/FlagOpen/FlagEmbedding.

Observational studies can yield clinically actionable evidence at scale, but executing them on real-world databases is open-ended and requires coherent decisions across cohort construction, analysis, and reporting. Prior evaluations of LLM agents emphasize isolated steps or single answers, missing the integrity and internal structure of the resulting evidence bundle. To address this gap, we introduce RWE-bench, a benchmark grounded in MIMIC-IV and derived from peer-reviewed observational studies. Each task provides the corresponding study protocol as the reference standard, requiring agents to execute experiments in a real database and iteratively generate tree-structured evidence bundles. We evaluate six LLMs (three open-source, three closed-source) under three agent scaffolds using both question-level correctness and end-to-end task metrics. Across 162 tasks, task success is low: the best agent reaches 39.9%, and the best open-source model reaches 30.4%. Agent scaffolds also matter substantially, causing over 30% variation in performance metrics. Furthermore, we implement an automated cohort evaluation method to rapidly localize errors and identify agent failure modes. Overall, the results highlight persistent limitations in agents' ability to produce end-to-end evidence bundles, and efficient validation remains an important direction for future work. Code and data are available at https://github.com/somewordstoolate/RWE-bench.

The needle-in-a-haystack (NIAH) test, which examines the ability to retrieve a piece of information (the "needle") from long distractor texts (the "haystack"), has been widely adopted to evaluate long-context language models (LMs). However, this simple retrieval-based test is indicative of only a superficial form of long-context understanding. To provide a more comprehensive evaluation of long-context LMs, we create a new synthetic benchmark RULER with flexible configurations for customized sequence length and task complexity. RULER expands upon the vanilla NIAH test to encompass variations with diverse types and quantities of needles. Moreover, RULER introduces new task categories multi-hop tracing and aggregation to test behaviors beyond searching from context. We evaluate ten long-context LMs with 13 representative tasks in RULER. Despite achieving nearly perfect accuracy in the vanilla NIAH test, all models exhibit large performance drops as the context length increases. While these models all claim context sizes of 32K tokens or greater, only four models (GPT-4, Command-R, Yi-34B, and Mixtral) can maintain satisfactory performance at the length of 32K. Our analysis of Yi-34B, which supports context length of 200K, reveals large room for improvement as we increase input length and task complexity. We open source RULER to spur comprehensive evaluation of long-context LMs.

Today's large language models (LLMs) can solve challenging question-answering tasks, and prompt engineering techniques, such as chain-of-thought (CoT), have gained attention for enhancing the explanation and correctness of outputs. Nevertheless, models require significant time to generate answers augmented with lengthy reasoning details. To address this issue, this paper analyzes the impact of output lengths on LLM inference pipelines and proposes novel metrics to evaluate them in terms of correct conciseness. It also examines the impact of controlling output length through a refined prompt engineering strategy, Constrained-CoT (CCoT), which encourages the model to limit output length. Experiments on pre-trained LLMs demonstrated the benefit of the proposed metrics and the effectiveness of CCoT across different models. For instance, constraining the reasoning of LLaMA2-70b to 100 words improves the accuracy from 36.01\% (CoT) to 41.07\% (CCoT) on the GSM8K dataset, while reducing the average output length by 28 words.

In traditional RAG framework, the basic retrieval units are normally short. The common retrievers like DPR normally work with 100-word Wikipedia paragraphs. Such a design forces the retriever to search over a large corpus to find the `needle' unit. In contrast, the readers only need to extract answers from the short retrieved units. Such an imbalanced `heavy' retriever and `light' reader design can lead to sub-optimal performance. In order to alleviate the imbalance, we propose a new framework LongRAG, consisting of a `long retriever' and a `long reader'. LongRAG processes the entire Wikipedia into 4K-token units, which is 30x longer than before. By increasing the unit size, we significantly reduce the total units from 22M to 700K. This significantly lowers the burden of retriever, which leads to a remarkable retrieval score: answer recall@1=71% on NQ (previously 52%) and answer recall@2=72% (previously 47%) on HotpotQA (full-wiki). Then we feed the top-k retrieved units (approx 30K tokens) to an existing long-context LLM to perform zero-shot answer extraction. Without requiring any training, LongRAG achieves an EM of 62.7% on NQ, which is the best known result. LongRAG also achieves 64.3% on HotpotQA (full-wiki), which is on par of the SoTA model. Our study offers insights into the future roadmap for combining RAG with long-context LLMs.

Evaluations of LLM systems routinely average over small prompt sets and report models as a ranked table. We ask how much confidence such a table deserves, using LLM-based prompt-structure inference as the case study: eight open model variants across five families and 8B to 675B parameters, caching disabled, 293 raw intermediate representations persisted. The measured phenomenon is unstable to begin with. Identical calls do not reliably recover identical structure, with mean node-set Jaccard from 0.39 to 0.96 and 72% of prompt-model cells never node-set-perfect. Auditing the evaluation weakens its conclusions further, and this is our main contribution. Under a joint cluster bootstrap over prompts, only the bottom of the ranking is firm: the two least reproducible models hold rank in 99% and 86% of replicates, the middle four in 27% to 48%, and the top two in 68% each, so the table identifies the worst model reliably but does not reliably identify the best. Two equally defensible rules for merging repeated campaigns change four of eight rows and move the study-wide headline by 7 percentage points. Checking the inferred structure against ground-truth annotations shows reproducibility cannot be read as accuracy. And four of the eight endpoints were withdrawn within ten weeks of measurement, so the study as specified can no longer be run. Small-sample LLM evaluations can therefore look far more definitive than their evidence supports. We recommend reporting rank stability, per-cell provenance, executed sensitivity comparisons, raw per-run outputs, and a measurement date alongside any ranking.

Large Language Models (LLMs) have achieved remarkable success in various natural language processing tasks, yet their ability to generate long-form content remains poorly understood and evaluated. Our analysis reveals that current LLMs struggle with length requirements and information density in long-text generation, with performance deteriorating as text length increases. To quantitively locate such a performance degradation and provide further insights on model development, we present LongEval, a benchmark that evaluates long-text generation through both direct and plan-based generation paradigms, inspired by cognitive and linguistic writing models. The comprehensive experiments in this work reveal interesting findings such as that while model size correlates with generation ability, the small-scale model (e.g., LongWriter), well-trained on long texts, has comparable performance. All code and datasets are released in https://github.com/Wusiwei0410/LongEval.

How does retrieval performance scale with pretraining FLOPs? We benchmark retrieval performance across LLM model sizes from 125 million parameters to 7 billion parameters pretrained on datasets ranging from 1 billion tokens to more than 2 trillion tokens. We find that retrieval performance on zero-shot BEIR tasks predictably scales with LLM size, training duration, and estimated FLOPs. We also show that In-Context Learning scores are strongly correlated with retrieval scores across retrieval tasks. Finally, we highlight the implications this has for the development of LLM-based retrievers.

Reinforcement Learning from Human Feedback relies on reward models to align large language models with human preferences. However, RLHF often suffers from reward hacking, wherein policy learning exploits flaws in the trained reward model to maximize reward scores without genuinely aligning with human preferences. A significant example of such reward hacking is length bias, where reward models usually favor longer responses irrespective of actual response quality. Previous works on length bias have notable limitations, these approaches either mitigate bias without characterizing the bias form, or simply assume a linear length-reward relation. To accurately model the intricate nature of length bias and facilitate more effective bias mitigation, we propose FiMi-RM (Bias Fitting to Mitigate Length Bias of Reward Model in RLHF), a framework that autonomously learns and corrects underlying bias patterns. Our approach consists of three stages: First, we train a standard reward model which inherently contains length bias. Next, we deploy a lightweight fitting model to explicitly capture the non-linear relation between length and reward. Finally, we incorporate this learned relation into the reward model to debias. Experimental results demonstrate that FiMi-RM achieves a more balanced length-reward distribution. Furthermore, when applied to alignment algorithms, our debiased reward model improves length-controlled win rate and reduces verbosity without compromising its performance.

Capability and efficiency are two key dimensions of reasoning in large language models (LLMs). Capability refers to the ability to solve a given problem correctly, whereas efficiency refers to the ability to do so with limited resources. When LLMs use Chain-of-Thought (CoT) reasoning to solve problems of controlled hardness, both the number of problems solved correctly and the number of tokens required to reach a correct answer depend on problem hardness and model size. However, how these factors jointly shape capability and efficiency remains poorly understood. Here, we use hierarchical Bayesian models to evaluate the capability and efficiency of LLMs from the DeepSeek-R1-Distill model family across four classes of arithmetic and algorithmic reasoning problems. At a fixed model size, the probability of correctly solving an instance decays approximately exponentially with instance size, our proxy for problem hardness. The decay scale grows sublinearly with model size, indicating that larger models are more capable, but that capability gains diminish with scale. Output length grows as a power law with instance size, which serves as a proxy for difficulty. However, the parameters of this power law do not vary systematically with model size, suggesting that larger models do not become more efficient. Together, these findings reveal potential limitations of naive scaling as a strategy for developing more capable AI systems: capability improves with diminishing returns, while efficiency shows little to no improvement.

Large language model (LLM) "answer engines" such as ChatGPT now send measurable referral traffic to the open web, and a practice analogous to search engine optimization, here called Answer Engine Optimization (AEO), has emerged. Public AEO success stories typically quote large raw growth multiples, but raw referral growth is confounded by the rapid platform-level growth of the answer engines themselves. We report a longitudinal field study on a single high-traffic domain (glasp.co) whose corpus of hundreds of thousands of YouTube question-and-answer pages received a defined bundle of AEO interventions in January 2026 (detailed in Section 4). Because the interventions were concentrated on one subset of the site, the untreated remainder of the same domain acts as a contemporaneous control that absorbs the platform tailwind. Using first-party analytics and server logs rather than probabilistic third-party estimators, we find: (1) raw growth is dominated by the platform tailwind: on monthly aggregates total ChatGPT referrals grew 5.7x while untreated pages on the same domain grew 3.5x over the same window; (2) an interrupted time-series model on the weekly treated/control ratio estimates a discrete, intervention-aligned level increase of 1.82x (95% CI 1.31-2.54, HAC p=0.001), robust across engagement-filtered traffic (2.27x) and alternative specifications; (3) however, a conservative placebo-in-time permutation test yields p=0.16, so the effect is suggestive, not conclusive, given a short and noisy pre-period; and (4) Google organic clicks to treated pages did not fall beyond the ambient site-wide trend and indexation was preserved, consistent with the SEO-protection rule. The methodological message, separating treatment from platform tailwind with an on-domain control, matters more than any single multiple, and implies that headline AEO multiples substantially overstate causal effect.

Scaling pre-training compute has proven effective for achieving mulitlinguality, but does the same hold for test-time scaling? In this work, we introduce MCLM, a multilingual math benchmark featuring competition-level problems in 55 languages. We test three test-time scaling methods-Outcome Reward Modeling (ORM), Process Reward Modeling (ORM), and Budget Forcing (BF)-on both Qwen2.5-1.5B Math and MR1-1.5B, a multilingual LLM we trained for extended reasoning. Our experiments show that using Qwen2.5-1.5B Math with ORM achieves a score of 35.8 on MCLM, while BF on MR1-1.5B attains 35.2. Although "thinking LLMs" have recently garnered significant attention, we find that their performance is comparable to traditional scaling methods like best-of-N once constrained to similar levels of inference FLOPs. Moreover, while BF yields a 20-point improvement on English AIME, it provides only a 1.94-point average gain across other languages-a pattern consistent across the other test-time scaling methods we studied-higlighting that test-time scaling may not generalize as effectively to multilingual tasks. To foster further research, we release MCLM, MR1-1.5B, and evaluation results.

Reinforcement learning with verifiable rewards (RLVR) has been shown to enhance the reasoning capabilities of large language models (LLMs), enabling the development of large reasoning models (LRMs). However, LRMs such as DeepSeek-R1 and OpenAI o1 often generate verbose responses containing redundant or irrelevant reasoning step-a phenomenon known as overthinking-which substantially increases computational costs. Prior efforts to mitigate this issue commonly incorporate length penalties into the reward function, but we find they frequently suffer from two critical issues: length collapse and training collapse, resulting in sub-optimal performance. To address them, we propose a pipeline for training a Conciseness Reward Model (CRM) that scores the conciseness of reasoning path. Additionally, we introduce a novel reward formulation named Conciseness Reward Function (CRF) with explicit dependency between the outcome reward and conciseness score, thereby fostering both more effective and more efficient reasoning. From a theoretical standpoint, we demonstrate the superiority of the new reward from the perspective of variance reduction and improved convergence properties. Besides, on the practical side, extensive experiments on five mathematical benchmark datasets demonstrate the method's effectiveness and token efficiency, which achieves an 8.1% accuracy improvement and a 19.9% reduction in response token length on Qwen2.5-7B. Furthermore, the method generalizes well to other LLMs including Llama and Mistral. The implementation code and datasets are publicly available for reproduction: https://anonymous.4open.science/r/CRM.

Large language models (LLMs) are now regularly asked to forecast real-world events, but comparisons are often difficult because models receive different information, use different tools, and are evaluated under different rules. This paper reports the completed AI World Cup benchmark, in which ten LLM-based assistants made a single pre-tournament forecast of the entire 2026 FIFA World Cup. Every submission used the same tournament snapshot, prompt, JSON schema, and scoring procedure. The forecasts covered group-stage scores, group rankings, the knockout bracket, final placings, confidence values, and short explanations. After all 104 matches had been played, GPT-5.5 Thinking finished first with 744 points, followed by GPT-5.5 with 717, Gemini with 699, and Qwen 3.7 with 687. GPT-5.5 Thinking was also the only model to select Spain, which defeated Argentina 1--0 in the final, as champion. The final ranking was driven mainly by knockout performance: total score was strongly correlated with knockout points (r=0.986), but showed little relationship with group-stage match points (r=0.055), group-standing points (r=-0.103), or their combined pre-knockout score (r=-0.054). Match-level accuracy produced a different ordering. Claude Sonnet 4.6 correctly predicted the largest number of group-stage outcomes (63.89\%) but placed sixth overall. Average self-reported confidence was also unrelated to either outcome accuracy (r=-0.060) or total score (r=-0.067). The results suggest that forecasting a complete tournament tests something different from predicting matches one at a time, while also showing how strongly a bracket-based leaderboard can depend on scoring design. The benchmark materials, raw responses, and scoring code are released to support replication and future extensions.

This paper models the availability of bikes at San Francisco Bay Area Bike Share stations using machine learning algorithms. Random Forest (RF) and Least-Squares Boosting (LSBoost) were used as univariate regression algorithms, and Partial Least-Squares Regression (PLSR) was applied as a multivariate regression algorithm. The univariate models were used to model the number of available bikes at each station. PLSR was applied to reduce the number of required prediction models and reflect the spatial correlation between stations in the network. Results clearly show that univariate models have lower error predictions than the multivariate model. However, the multivariate model results are reasonable for networks with a relatively large number of spatially correlated stations. Results also show that station neighbors and the prediction horizon time are significant predictors. The most effective prediction horizon time that produced the least prediction error was 15 minutes.

Agent Skills package procedural instructions and checks for use by general-purpose agents, but loading a skill does not guarantee that every requirement remains active throughout a long tool-using trajectory. We study this problem in a production-derived, white-box code-audit workflow. Holding the task and 24 artifact checks fixed, we vary the surrounding context and classify where failures first become visible: lost requirements, editing drift, failed checking, or non-agent evaluator/runtime failures. Codex with gpt-5.4-mini passes 8/10 runs in a 10,991-character clean context but only 3/10 in both a 299,140-character relevant context and an equal-length irrelevant context. This 50-percentage-point difference is large but remains trend-level under two-sided Fisher tests (p = 0.0698). Requirement coverage nevertheless stays above 92% in both long conditions, showing that a few omissions can invalidate an otherwise complete artifact. A second task passes all clean and long runs, so the evidence does not support a universal context-length threshold. A detailed external checklist passes 10/10 runs, compared with 5/10 for a generic self-check (p = 0.0325). Coding-agent scaffolds may help by selecting a smaller working set, but they do not eliminate failures. We do not introduce context rot or a new general monitoring method; we provide a bounded failure classification and empirical case study for white-box code auditing.

Understanding long videos, ranging from tens of minutes to several hours, presents unique challenges in video comprehension. Despite the increasing importance of long-form video content, existing benchmarks primarily focus on shorter clips. To address this gap, we introduce InfiniBench a comprehensive benchmark for very long video understanding which presents 1)The longest video duration, averaging 76.34 minutes; 2) The largest number of question-answer pairs, 108.2K; 3) Diversity in questions that examine nine different skills and include both multiple-choice questions and open-ended questions; 4) Humancentric, as the video sources come from movies and daily TV shows, with specific human-level question designs such as Movie Spoiler Questions that require critical thinking and comprehensive understanding. Using InfiniBench, we comprehensively evaluate existing Large MultiModality Models (LMMs) on each skill, including the commercial model Gemini 1.5 Flash and the open-source models. The evaluation shows significant challenges in our benchmark.Our results show that the best AI models such Gemini struggles to perform well with 42.72% average accuracy and 2.71 out of 5 average score. We hope this benchmark will stimulate the LMMs community towards long video and human-level understanding. Our benchmark can be accessed at https://vision-cair.github.io/InfiniBench/

Transformer language models typically operate with a fixed-length context window, which has grown in step with large-scale pretraining datasets. In the BabyLM Challenge, however, many past submissions have defaulted to using much shorter sequence lengths. We examine the impact of sequence length on BabyLM pretraining, to answer the simple question: what sequence length should we be using when training Baby LMs? Using 100M-word training data and fixed compute budgets, we compare 125M-parameter Mamba and OPT models, finding that although longer is often better, the optimal length depends on both task and architecture. Shorter sequences are sufficient for grammatical generalization tasks whereas longer contexts benefit morphological analogical reasoning tasks.

The rapid adoption of AI coding assistants and autonomous agentic development systems has coincided with major changes in the pace and structure of open-source software engineering. Yet empirical longitudinal evidence of these changes at the team level remains limited. We present a descriptive longitudinal analysis of seven engineering metrics: pull request (PR) throughput, cycle time, contributor diversity, PR comment density, merge rate, new-author participation, and PR size. Metrics were computed from all merged PRs in two high-velocity AI infrastructure repositories, vLLM (February 2023-June 2026; 18,290 PRs) and SGLang (January 2024-June 2026; 14,938 PRs). We segment development into four eras aligned with major changes in AI-assisted software development and examine human- and bot-authored activities. Both projects show substantial increases in development velocity and AI-developer collaboration signals. PR throughput increased 21x in vLLM and 17.9x in SGLang, while bot-authored PRs accounted for less than 0.2% of this growth, indicating that the increase was overwhelmingly human-driven. In the latest era, median cycle time was 1.04 days for vLLM and 0.62 days for SGLang, while P90 cycle times reached 16.8 and 14.3 days, respectively. Monthly unique authors increased steadily in both projects, suggesting broader contributor participation. PR comment density increased 4.2x in vLLM and 3.8x in SGLang, with bot comments contributing an estimated 15-20% of the increase. In contrast, PR size remained relatively stable across eras. Overall, AI-assisted development is associated with higher throughput, broader contributor participation, and increased AI-developer collaboration signals in high-velocity open-source software development.

LLMs are not generally able to adjust the length of their outputs based on strict length requirements, a capability that would improve their usefulness in applications that require adherence to diverse user and system requirements. We present an approach to train LLMs to acquire this capability by augmenting existing data and applying existing fine-tuning techniques, which we compare based on the trained models' adherence to the length requirement and overall response quality relative to the baseline model. Our results demonstrate that these techniques can be successfully applied to train LLMs to adhere to length requirements, with the trained models generating texts which better align to the length requirements. Our results indicate that our method may change the response quality when using training data that was not generated by the baseline model. This allows simultaneous alignment to another training objective in certain scenarios, but is undesirable otherwise. Training on a dataset containing the model's own responses eliminates this issue.

Loop Engineering is emerging as a practice for organizing development work around coding agents. Instead of writing each prompt by hand, practitioners design loops that monitor progress, assign work, run checks, and decide what the agent should do next. Even with a capable coding agent, a loop may trust a stale progress note, skip needed verification, spend its budget in the wrong direction, or stop before the task is safe to submit. Yet the final outcome of one end-to-end run cannot tell whether success or failure reflects the loop's guidance or the coding agent's ability to carry out the task. We introduce LoopArena, a benchmark for evaluating how well one model can guide a separate coding agent through a long-running task. The model under evaluation is the Controller: after each coding round, it receives a structured summary of the run and instructs a separate, fixed coding agent, the Worker, on what to do or verify next, or decides whether to stop. LoopArena evaluates this ability in three complementary settings that differ in execution scope and cost. Type I scores next-step Loop Contract selection through execution-validated questions without running the Worker at evaluation time. Type II executes repeated control over a selected slice of a full task, while Type III evaluates the paired full task from its original state. On full tasks, the best observed Strict Success Rate is 24.69\%, leaving substantial room for improvement in long-horizon loop control. Across Controllers, the paired reduction in estimated inference cost averages 64.4\%, and Type II produces a similar ordering under the main Core criterion (Spearman's \(ρ=0.9747\)). We release the benchmark data and evaluation code at https://github.com/AMAP-ML/LoopArena .

There is a consensus that instruction fine-tuning of LLMs requires high-quality data, but what are they? LIMA (NeurIPS 2023) and AlpaGasus (ICLR 2024) are state-of-the-art methods for selecting such high-quality examples, either via manual curation or using GPT-3.5-Turbo as a quality scorer. We show that the extremely simple baseline of selecting the 1,000 instructions with longest responses from standard datasets can consistently outperform these sophisticated methods according to GPT-4 and PaLM-2 as judges, while remaining competitive on the OpenLLM benchmarks that test factual knowledge. We demonstrate this for several state-of-the-art LLMs (Llama-2-7B, Llama-2-13B, and Mistral-7B) and datasets (Alpaca-52k and Evol-Instruct-70k). In addition, a lightweight refinement of such long instructions can further improve the abilities of the fine-tuned LLMs, and allows us to obtain the 2nd highest-ranked Llama-2-7B-based model on AlpacaEval 2.0 while training on only 1,000 examples and no extra preference data. We also conduct a thorough analysis of our models to ensure that their enhanced performance is not simply due to GPT-4's preference for longer responses, thus ruling out any artificial improvement. In conclusion, our findings suggest that fine-tuning on the longest instructions should be the default baseline for any research on instruction fine-tuning.

The growth rate of Av(1324) is the last unknown Stanley-Wilf limit of a length-four pattern. Since 2020 the best rigorous lower bound has been the 10.271012 of Bevan, Brignall, Elvey Price and Pantone; we raise it to 10.412263. Their construction lays 1324-avoiders along an infinite staircase of two-cell blocks alternating with single connecting cells, and forbids a 1324 by a local condition on how the points of a block interleave with the components of the cell beside it. Certain points of a block, its leaves, may be exempted from that condition. They exempt them on one axis only, and report that they could not count the possibilities when both are exempted. This paper exempts them on both axes and counts the result. The second relaxation is their own argument read in the other axis, and the Harris correlation inequality bounds the joint count below by the product of the two counts taken separately, at the cost of one factor of the component generating function.

The ability of Large Language Models (LLMs) to generate structured outputs, such as JSON, is crucial for their use in Compound AI Systems. However, evaluating and improving this capability remains challenging. In this work, we introduce StructuredRAG, a benchmark of six tasks designed to assess LLMs' proficiency in following response format instructions. We evaluate two state-of-the-art LLMs, Gemini 1.5 Pro and Llama 3 8B-instruct with 4-bit quantization using two distinct prompting strategies. We introduce these prompting strategies as f-String and Follow the Format (FF) prompting. Across 24 experiments, we find an average success rate of 82.55%. We further find a high variance in performance across tasks, models, and prompting strategies with success rates ranging from 0 to 100%. We find that Llama 3 8B-instruct often performs competitively with Gemini 1.5 Pro. We observe that task complexity significantly influences performance, with tasks involving lists or composite object outputs proving more challenging. Our findings highlight the need for further research into improving the reliability and consistency of structured output generation in LLMs. We have open-sourced our experimental code and results at github.com/weaviate/structured-rag.

Large vision-language models (VLMs) can recognize what happens in video but fail to count how many times. We introduce PushupBench, 446 long-form clips (avg. 36.7s) for evaluating repetition counting. The best frontier model achieves 42.1\% exact accuracy; open-source 4B models score sim6\%, matching supervised baselines. We show that accuracy alone misleads -- weaker models exploit the modal count rather than reason temporally. Fine-tuning on counting with 1k samples transfers to general video understanding: MVBench (+2.15), PerceptionTest (+1.88), TVBench (+4.54), suggesting counting is a proxy for broader temporal reasoning.PushupBench incorporated in lmms-eval (https://github.com/EvolvingLMMs-Lab/lmms-eval/pull/1262) and hosted on (pushupbench.com/)

Reinforcement Learning with Verifiable Rewards (RLVR) for large language models (LLMs) has achieved remarkable progress in enhancing LLMs' reasoning capabilities on tasks with clear correctness criteria, such as mathematical reasoning tasks. Several training metrics, such as entropy or response length, have been observed to correlate with different reasoning behaviors in reinforcement learning. Prior approaches incorporate such priors through reward or advantage shaping, which often relies on hand-crafted penalties and preferences (e.g., higher-is-better or lower-is-better). However, without careful hyperparameter tuning, these directional priors can be overly biased and may lead to failure. To this end, we introduce Conditional advANtage estimatiON (CANON), amplifying the impact of the target metric without presuming its direction. Specifically, CANON regroups the sampled responses into two groups based on the higher or lower value of a target metric, measures which metric trend contributes to better performance through inter-group comparison, and identifies the better response within the same group. In summary, CANON based on entropy consistently outperforms prior methods across three LLMs on both math reasoning and high-complexity logic tasks. When applied to response length, CANON further improves token efficiency, yielding a more favorable Pareto frontier in the performance-cost trade-off.

Recent applications of Reinforcement Learning with Verifiable Rewards (RLVR) to Large Language Models (LLMs) and Vision-Language Models (VLMs) have demonstrated significant success in enhancing reasoning capabilities for complex tasks. During RLVR training, an increase in response length is often regarded as a key factor contributing to the growth of reasoning ability. However, the patterns of change in response length vary significantly across different RLVR algorithms during the training process. To provide a fundamental explanation for these variations, this paper conducts an in-depth analysis of the components of mainstream RLVR algorithms. We present a theoretical analysis of the factors influencing response length and validate our theory through extensive experimentation. Building upon these theoretical findings, we propose the Length-Unbiased Sequence Policy Optimization (LUSPO) algorithm. Specifically, we rectify the length bias inherent in Group Sequence Policy Optimization (GSPO), rendering its loss function unbiased with respect to response length and thereby resolving the issue of response length collapse. We conduct extensive experiments across mathematical reasoning benchmarks and multimodal reasoning scenarios, where LUSPO consistently achieves superior performance. Empirical results demonstrate that LUSPO represents a novel, state-of-the-art optimization strategy compared to existing methods such as GRPO and GSPO.

Prompt optimization in compound AI systems is statistically indistinguishable from a coin flip: across 72 optimization runs on Claude Haiku 4.5 (6 methods times 4 tasks times 3 repeats), 49% score below zero-shot; on Amazon Nova Lite, the failure rate is even higher. Yet on one task, all six methods improve over zero-shot by up to +6.8 points. What distinguishes success from failure? We investigate with 18,000 grid evaluations and 144 optimization runs, testing two assumptions behind end-to-end optimization tools like TextGrad and DSPy, in the order they must be answered: (A) agent prompts interact, requiring joint rather than independent optimization, and (B) individual prompts are worth optimizing at all. Interaction effects are never significant (p > 0.52, all F < 1.0), and optimization helps only when the task has exploitable output structure: a format the model can produce but does not default to. We further give a mechanistic account: instruction-tuning compresses input phrasing into a narrow output distribution, eliminating the very phrasing-sensitivity that joint optimization assumes. We provide a two-stage diagnostic: an \$80 ANOVA pre-test for agent coupling, and a 10-minute headroom test that predicts whether optimization is worthwhile, turning a coin flip into an informed decision.

Accurately evaluating machine-translated text remains a long-standing challenge, particularly for long documents. Recent work has shown that large language models (LLMs) can serve as reliable and interpretable sentence-level translation evaluators via MQM error span annotations. With modern LLMs supporting larger context windows, a natural question arises: can we feed entire document translations into an LLM for quality assessment? Ideally, evaluation should be invariant to text length, producing consistent error spans regardless of input granularity. However, our analysis shows that text length significantly impacts evaluation: longer texts lead to fewer error spans and reduced system ranking accuracy. To address this limitation, we evaluate several strategies, including granularity-aligned prompting, Focus Sentence Prompting (FSP), and a fine-tuning approach to better align LLMs with the evaluation task. The latter two methods largely mitigate this length bias, making LLMs more reliable for long-form translation evaluation.

Recently, there has been growing interest in extending the context length of instruction-following models in order to effectively process single-turn long input (e.g. summarizing a paper) and conversations with more extensive histories. While proprietary models such as GPT-4 and Claude have demonstrated considerable advancements in handling tens of thousands of tokens of context, open-sourced models are still in the early stages of experimentation. It also remains unclear whether developing these long context models can offer substantial gains on practical downstream tasks over retrieval-based methods or models simply trained on chunked contexts. To address this challenge, we propose to institute standardized evaluation for long context language models. Concretely, we develop L-Eval which contains 411 long documents and over 2,000 query-response pairs manually annotated and checked by the authors encompassing areas such as law, finance, school lectures, lengthy conversations, news, long-form novels, and meetings. L-Eval also adopts diverse evaluation methods and instruction styles, enabling a more reliable assessment of Long Context Language Models (LCLMs). Our findings indicate that while open-source models typically lag behind their commercial counterparts, they still exhibit impressive performance. LLaMA2 achieves the best results (win 45\% vs turbo-16k) on open-ended tasks with only 4k context length and ChatGLM2 achieves the best results on closed-ended tasks with 8k input tokens. We release our new evaluation suite, code, and all generation results including predictions from all open-sourced LCLMs, GPT4-32k, Cluade-100k at {https://github.com/OpenLMLab/LEval}.

Retrieval Augmented Generation (RAG) has emerged as a crucial technique for enhancing the accuracy of Large Language Models (LLMs) by incorporating external information. With the advent of LLMs that support increasingly longer context lengths, there is a growing interest in understanding how these models perform in RAG scenarios. Can these new long context models improve RAG performance? This paper presents a comprehensive study of the impact of increased context length on RAG performance across 20 popular open source and commercial LLMs. We ran RAG workflows while varying the total context length from 2,000 to 128,000 tokens (and 2 million tokens when possible) on three domain-specific datasets, and report key insights on the benefits and limitations of long context in RAG applications. Our findings reveal that while retrieving more documents can improve performance, only a handful of the most recent state of the art LLMs can maintain consistent accuracy at long context above 64k tokens. We also identify distinct failure modes in long context scenarios, suggesting areas for future research.

Despite their recent successes, Transformer-based large language models show surprising failure modes. A well-known example of such failure modes is their inability to length-generalize: solving problem instances at inference time that are longer than those seen during training. In this work, we further explore the root cause of this failure by performing a detailed analysis of model behaviors on the simple parity task. Our analysis suggests that length generalization failures are intricately related to a model's inability to perform random memory accesses within its context window. We present supporting evidence for this hypothesis by demonstrating the effectiveness of methodologies that circumvent the need for indexing or that enable random token access indirectly, through content-based addressing. We further show where and how the failure to perform random memory access manifests through attention map visualizations.

Prior work on a social web highlighter located individuality in selection -- which documents a person chooses to highlight -- but measured it cross-sectionally. We ask the temporal question: is a reader's selection signature a trait or a state? We freeze each reader's first six months of highlighting as a profile and track its own-vs-other advantage on their later selections at growing gaps (to 24+ months), with negatives drawn from the same calendar era -- so supply drift cannot masquerade as personal drift -- at a coarse global level and at a fine level whose negatives and controls come from the reader's own interest neighborhood; the anchor cell reproduces the prior cross-sectional level (+0.188 vs +0.169), validating the harness. Four results. Within the same users, the fine-layer advantage shows no statistically detectable paired decline at any horizon (6-12 month retention R = 1.00 [0.85, 1.18], n = 212; the farthest bin is compatible with a modest decline; the only contrast whose interval excludes zero is the coarse layer at 12-24 months, about 13%). The signal is not reducible to repeated domains (~90% survives excluding all profile sources). Within-person drift is slow (a recent-half profile beats the old half by +0.042). Prospectively, personal profiles -- even one built from a reader's earliest documents, median 20 months before evaluation -- rank their next reads at roughly 3x the AP of every simple non-personal prior tested. We use "trait" operationally (a stable signature under continued engagement); the scope is heavy, long-tenured readers of one platform, and exposure is not separable from choice.

Pretrained language models (LMs) perform well on many tasks even when learning from a few examples, but prior work uses many held-out examples to tune various aspects of learning, such as hyperparameters, training objectives, and natural language templates ("prompts"). Here, we evaluate the few-shot ability of LMs when such held-out examples are unavailable, a setting we call true few-shot learning. We test two model selection criteria, cross-validation and minimum description length, for choosing LM prompts and hyperparameters in the true few-shot setting. On average, both marginally outperform random selection and greatly underperform selection based on held-out examples. Moreover, selection criteria often prefer models that perform significantly worse than randomly-selected ones. We find similar results even when taking into account our uncertainty in a model's true performance during selection, as well as when varying the amount of computation and number of examples used for selection. Overall, our findings suggest that prior work significantly overestimated the true few-shot ability of LMs given the difficulty of few-shot model selection.

Large language models (LLMs) are increasingly used for tasks that require complex reasoning. Most benchmarks focus on final outcomes but overlook the intermediate reasoning steps - such as planning, revision, and decision making under resource constraints. We argue that measuring these internal processes is essential for understanding model behavior and improving reliability. We propose using strategic games as a natural evaluation environment: closed, rule-based systems with clear states, limited resources, and automatic feedback. We introduce a framework that evaluates LLMs along three core dimensions: planning, revision, and resource-constrained decision making. To operationalize this, we define metrics beyond win rate, including overcorrection risk rate, correction success rate, improvement slope, and over-budget ratio. In 4320 adversarial rounds across 12 leading models, ChatGPT-o3-mini achieves the top composite score, with a win rate of 74.7 percent, a correction success rate of 78.6 percent, and an improvement slope of 0.041. By contrast, Qwen-Plus, despite an overcorrection risk rate of 81.6 percent, wins only 25.6 percent of its matches - primarily due to excessive resource use. We also observe a negative correlation between overcorrection risk rate and correction success rate (Pearson r = -0.51, p = 0.093), suggesting that more frequent edits do not always improve outcomes. Our findings highlight the value of assessing not only what LLMs decide but how they arrive at those decisions

Background: Recent advancements in large language models (LLMs) offer potential benefits in healthcare, particularly in processing extensive patient records. However, existing benchmarks do not fully assess LLMs' capability in handling real-world, lengthy clinical data. Methods: We present the LongHealth benchmark, comprising 20 detailed fictional patient cases across various diseases, with each case containing 5,090 to 6,754 words. The benchmark challenges LLMs with 400 multiple-choice questions in three categories: information extraction, negation, and sorting, challenging LLMs to extract and interpret information from large clinical documents. Results: We evaluated nine open-source LLMs with a minimum of 16,000 tokens and also included OpenAI's proprietary and cost-efficient GPT-3.5 Turbo for comparison. The highest accuracy was observed for Mixtral-8x7B-Instruct-v0.1, particularly in tasks focused on information retrieval from single and multiple patient documents. However, all models struggled significantly in tasks requiring the identification of missing information, highlighting a critical area for improvement in clinical data interpretation. Conclusion: While LLMs show considerable potential for processing long clinical documents, their current accuracy levels are insufficient for reliable clinical use, especially in scenarios requiring the identification of missing information. The LongHealth benchmark provides a more realistic assessment of LLMs in a healthcare setting and highlights the need for further model refinement for safe and effective clinical application. We make the benchmark and evaluation code publicly available.

Survival analysis is the problem of estimating probability distributions for future event times, which can be seen as a problem in uncertainty quantification. Although there are fundamental theories on strictly proper scoring rules for uncertainty quantification, little is known about those for survival analysis. In this paper, we investigate extensions of four major strictly proper scoring rules for survival analysis and we prove that these extensions are proper under certain conditions, which arise from the discretization of the estimation of probability distributions. We also compare the estimation performances of these extended scoring rules by using real datasets, and the extensions of the logarithmic score and the Brier score performed the best.

Length generalization failure problem, namely the large language model (LLM) fails to generalize to texts longer than its maximum training length, greatly restricts the application of LLM in the scenarios with streaming long inputs. To address this problem, the existing methods either require substantial costs or introduce precision loss. In this paper, we empirically find that the accuracy of the LLM's prediction is highly correlated to its certainty. Based on this, we propose an efficient training free framework, named XL3M (it means extra-long large language model), which enables the LLMs trained on short sequences to reason extremely long sequence without any further training or fine-tuning. Under the XL3M framework, the input context will be firstly decomposed into multiple short sub-contexts, where each sub-context contains an independent segment and a common ``question'' which is a few tokens from the end of the original context. Then XL3M gives a method to measure the relevance between each segment and the ``question'', and constructs a concise key context by splicing all the relevant segments in chronological order. The key context is further used instead of the original context to complete the inference task. Evaluations on comprehensive benchmarks show the superiority of XL3M. Using our framework, a Llama2-7B model is able to reason 20M long sequences on an 8-card Huawei Ascend 910B NPU machine with 64GB memory per card.

Probing neural models for the ability to perform downstream tasks using their activation patterns is often used to localize what parts of the network specialize in performing what tasks. However, little work addressed potential mediating factors in such comparisons. As a test-case mediating factor, we consider the prediction's context length, namely the length of the span whose processing is minimally required to perform the prediction. We show that not controlling for context length may lead to contradictory conclusions as to the localization patterns of the network, depending on the distribution of the probing dataset. Indeed, when probing BERT with seven tasks, we find that it is possible to get 196 different rankings between them when manipulating the distribution of context lengths in the probing dataset. We conclude by presenting best practices for conducting such comparisons in the future.

Practitioners make three prompt-design decisions with almost no controlled evidence behind them: how to format instructions and context (markdown, plain text, prose, or tabular), how many simultaneous instructions a system prompt can carry before compliance degrades, and how much context a model can hold before recall and honesty degrade. We report two controlled experiments crossing all three factors on one held, contamination-free synthetic corpus (the "Book of Veyra," 8,780 uniquely-named entities, deterministically regenerable from a fixed seed), evaluated across five models. Experiment 1 (960 calls/model) measures instruction-following decay as rule count N grows from 10 to 160, crossed with four formats and system-prompt vs. user-turn placement. Perfect-response rate collapses to zero by N=80 for every model, format, and placement. Placement produces effects at least as large as format at N=160 in most models, but the direction is model-specific. No model shows a reliable markdown advantage; one 35B model favors plain text instead. Experiment 2 (5,520 calls/model) measures recall accuracy, false-premise sycophancy, and absent-fact fabrication across a 2k-to-512k-token context ladder in the same four formats. Recall stays near ceiling through 64-128k tokens, then degrades sharply and format-dependently: one model's accuracy spread reaches 48 points at 128k tokens. Fabrication never occurs (0/5,760 probes), and sycophancy stays negligible (<=8.3%). What rises sharply near each model's context ceiling is outright refusal to answer (0% to 79-90%), distinct from sycophancy or fabrication. Neither pre-registered format ordering holds, and token overhead (+22% to +37% over plain text) further changes which format is preferable where accuracy spread is genuine. We release the full harness, corpus generator, and raw results (VeyraBench): https://github.com/iNetanel/veyrabench

We introduce a general, flexible, parametric survival modelling framework which encompasses key shapes of hazard function (constant, increasing, decreasing, up-then-down, down-then-up), various common survival distributions (log-logistic, Burr type XII, Weibull, Gompertz), and includes defective distributions (i.e., cure models). This generality is achieved using four basic distributional parameters: two scale-type parameters and two shape parameters. Generalising to covariate dependence, the scale-type regression components correspond to accelerated failure time (AFT) and proportional hazards (PH) models. Therefore, this general formulation unifies the most popular survival models which allows us to consider the practical value of possible modelling choices for survival data. Furthermore, in line with our proposed flexible baseline distribution, we advocate the use of multi-parameter regression in which more than one distributional parameter depends on covariates - rather than the usual convention of having a single covariate-dependent (scale) parameter. While many choices are available, we suggest introducing covariates through just one or other of the two scale parameters, which covers AFT and PH models, in combination with a `power' shape parameter, which allows for more complex non-AFT/non-PH effects, while the other shape parameter remains covariate-independent, and handles automatic selection of the baseline distribution. We explore inferential issues in simulations, both with and without a covariate, with particular focus on evidence concerning the need, or otherwise, to include both AFT and PH parameters. We illustrate the efficacy of our modelling framework by investigating differences between treatment groups using data from a lung cancer study and a melanoma study. Censoring is accommodated throughout.

What makes a good Large Language Model (LLM)? That it performs well on the relevant benchmarks -- which hopefully measure, with some validity, the presence of capabilities that are also challenged in real application. But what makes the model perform well? What gives a model its abilities? We take a recently introduced type of benchmark that is meant to challenge capabilities in a goal-directed, agentive context through self-play of conversational games, and analyse how performance develops as a function of model characteristics like number of parameters, or type of training. We find that while there is a clear relationship between number of parameters and performance, there is still a wide spread of performance points within a given size bracket, which is to be accounted for by training parameters such as fine-tuning data quality and method. From a more practical angle, we also find a certain degree of unpredictability about performance across access methods, possible due to unexposed sampling parameters, and a, very welcome, performance stability against at least moderate weight quantisation during inference.

In recent years, there have been remarkable advancements in the performance of Transformer-based Large Language Models (LLMs) across various domains. As these LLMs are deployed for increasingly complex tasks, they often face the needs to conduct longer reasoning processes or understanding larger contexts. In these situations, the length generalization failure of LLMs on long sequences become more prominent. Most pre-training schemes truncate training sequences to a fixed length (such as 2048 for LLaMa). LLMs often struggle to generate fluent texts, let alone carry out downstream tasks, after longer contexts, even with relative positional encoding which is designed to cope with this problem. Common solutions such as finetuning on longer corpora often involves daunting hardware and time costs and requires careful training process design. To more efficiently leverage the generation capacity of existing LLMs, we theoretically and empirically investigate the main out-of-distribution (OOD) factors contributing to this problem. Inspired by this diagnosis, we propose a simple yet effective solution for on-the-fly length generalization, LM-Infinite, which involves only a Lambda-shaped attention mask and a distance limit while requiring no parameter updates or learning. We find it applicable to a variety of LLMs using relative-position encoding methods. LM-Infinite is computational efficient with O(n) time and space, and demonstrates consistent fluency and generation quality to as long as 32k tokens on ArXiv and OpenWebText2 datasets, with 2.72x decoding speedup. On downstream task such as passkey retrieval, it continues to work on inputs much longer than training lengths where vanilla models fail immediately.

The ACM Recommender Systems Challenge 2018 focused on the task of automatic music playlist continuation, which is a form of the more general task of sequential recommendation. Given a playlist of arbitrary length with some additional meta-data, the task was to recommend up to 500 tracks that fit the target characteristics of the original playlist. For the RecSys Challenge, Spotify released a dataset of one million user-generated playlists. Participants could compete in two tracks, i.e., main and creative tracks. Participants in the main track were only allowed to use the provided training set, however, in the creative track, the use of external public sources was permitted. In total, 113 teams submitted 1,228 runs to the main track; 33 teams submitted 239 runs to the creative track. The highest performing team in the main track achieved an R-precision of 0.2241, an NDCG of 0.3946, and an average number of recommended songs clicks of 1.784. In the creative track, an R-precision of 0.2233, an NDCG of 0.3939, and a click rate of 1.785 was obtained by the best team. This article provides an overview of the challenge, including motivation, task definition, dataset description, and evaluation. We further report and analyze the results obtained by the top performing teams in each track and explore the approaches taken by the winners. We finally summarize our key findings, discuss generalizability of approaches and results to domains other than music, and list the open avenues and possible future directions in the area of automatic playlist continuation.

Aggregate metacognitive quality scores mask within-model variation across MMLU benchmark domains. We administered 1,500 MMLU items (250 per domain, under an a priori six-domain grouping) to 33 frontier LLMs from eight model families and computed Type-2 AUROC per model-domain cell using verbalized confidence (0-100). Total observations: 47,151. Every model with above-chance aggregate monitoring showed non-trivial domain-level variation. Applied/Professional knowledge was reliably the easiest benchmark domain to monitor (mean AUROC = .742, ranked top-2 in 21 of 33 models); Formal Reasoning and Natural Science were reliably the hardest (one of the two ranked bottom-2 in 27 of 33 models). The three middle domains were statistically indistinguishable (Kendall's W = .164). A subject-level coherence analysis (within-domain similarity ratio = 0.95) confirms the six-domain grouping is a pragmatic benchmark taxonomy, not a validated latent construct. Within-family profile-shape clustering is significant for Anthropic, Google-Gemini, and Qwen (permutation p < .0001) but not DeepSeek, Google-Gemma, or OpenAI. Gemma 4 31B showed a +.202 AUROC improvement over Gemma 3 27B. Three models classified Invalid on binary KEEP/WITHDRAW probes produced normal profiles under verbalized confidence, confirming probe-format specificity. Bootstrap 95% CIs on 198 cells have median width .199. Split-half aggregate stability r = .893; profile-level split-half is weaker (grand median r = .184). These results show stable benchmark-domain variation obscured by aggregate metrics, and support benchmark-stage domain screening as a step before deployment in specific application areas.

A growing literature argues that progress in human longevity is slowing, and that this slowdown may signal an approaching biological limit to lifespan. The argument rests on visible flattening of cumulative life expectancy records, evaluated through linear regressions or through comparisons of average gains across decades. These summaries, however, provide only weak evidence for slowing improvements, because they cannot distinguish a true from an apparent slowdown. An apparent slowdown may arise due to periods of low variance around unchanged mean annual gains. Applying a Bayesian change-point framework to annual gains, both at the best-practice frontier and separately for each record-holding country, we find no evidence that the expected annual gain has changed across nearly two centuries, but strong evidence that its variance has moved through distinct historical regimes, with progress most regular in recent decades. The recent flattening is consistent with a low-variance regime and provides little evidence for a decline in the underlying mean pace of improvement. Claims that the best-practice frontier signals a biological ceiling should therefore not rest on the recent bend in the cumulative record alone. The frontier is better read as an empirical benchmark, the highest country-specific survival achieved under observed historical conditions.

Large language models cannot estimate how long their own tasks take. We investigate this limitation through four experiments across 68 tasks and four model families. Pre-task estimates overshoot actual duration by 4--7times (p < 0.001), with models predicting human-scale minutes for tasks completing in seconds. Relative ordering fares no better: on task pairs designed to expose heuristic reliance, models score at or below chance (GPT-5: 18\% on counter-intuitive pairs, p = 0.033), systematically failing when complexity labels mislead. Post-hoc recall is disconnected from reality -- estimates diverge from actuals by an order of magnitude in either direction. These failures persist in multi-step agentic settings, with errors of 5--10times. The models possess propositional knowledge about duration from training but lack experiential grounding in their own inference time, with practical implications for agent scheduling, planning and time-critical scenarios.

In this study, we approached the Hit Song Prediction problem, which aims to predict which songs will become Billboard hits. We gathered a dataset of nearly 18500 hit and non-hit songs and extracted their audio features using the Spotify Web API. We test four machine-learning models on our dataset. We were able to predict the Billboard success of a song with approximately 86\% accuracy. The most succesful algorithms were Random Forest and Support Vector Machine.

Many NLP researchers rely on free computational services, such as Google Colab, to fine-tune their Transformer models, causing a limitation for hyperparameter optimization (HPO) in long-text classification due to the method having quadratic complexity and needing a bigger resource. In Indonesian, only a few works were found on long-text classification using Transformers. Most only use a small amount of data and do not report any HPO. In this study, using 18k news articles, we investigate which pretrained models are recommended to use based on the output length of the tokenizer. We then compare some hacks to shorten and enrich the sequences, which are the removals of stopwords, punctuation, low-frequency words, and recurring words. To get a fair comparison, we propose and run an efficient and dynamic HPO procedure that can be done gradually on a limited resource and does not require a long-running optimization library. Using the best hack found, we then compare 512, 256, and 128 tokens length. We find that removing stopwords while keeping punctuation and low-frequency words is the best hack. Some of our setups manage to outperform taking 512 first tokens using a smaller 128 or 256 first tokens which manage to represent the same information while requiring less computational resources. The findings could help developers to efficiently pursue optimal performance of the models using limited resources.

Context: Motivation is known to improve performance. In software development in particular, there has been considerable interest in the motivation of contributors to open source. Objective: We identify 11 motivators from the literature (enjoying programming, ownership of code, learning, self use, etc.), and evaluate their relative effect on motivation. Since motivation is an internal subjective feeling, we also analyze the validity of the answers. Method: We conducted a survey with 66 questions on motivation which was completed by 521 developers. Most of the questions used an 11 point scale. We evaluated the validity of the answers validity by comparing related questions, comparing to actual behavior on GitHub, and comparison with the same developer in a follow up survey. Results: Validity problems include moderate correlations between answers to related questions, as well as self promotion and mistakes in the answers. Despite these problems, predictive analysis, investigating how diverse motivators influence the probability of high motivation, provided valuable insights. The correlations between the different motivators are low, implying their independence. High values in all 11 motivators predict increased probability of high motivation. In addition, improvement analysis shows that an increase in most motivators predicts an increase in general motivation.

The Growth-at-Risk (GaR) framework has garnered attention in recent econometric literature, yet current approaches implicitly assume a constant Pareto exponent. We introduce novel and robust econometrics to estimate the tails of GaR based on a rigorous theoretical framework and establish validity and effectiveness. Simulations demonstrate consistent outperformance relative to existing alternatives in terms of predictive accuracy. We perform a long-term GaR analysis that provides accurate and insightful predictions, effectively capturing financial anomalies better than current methods.

Large Language Models (LLMs) have shown promise in accelerating the scientific research pipeline. A key capability for this process is the ability to generate novel research ideas, and prior studies have found settings in which LLM-generated research ideas were judged as more novel than human-expert ideas. However, a good idea should not simply appear to be novel, it should also result in better research after being executed. To test whether AI-generated ideas lead to better research outcomes, we conduct an execution study by recruiting 43 expert researchers to execute randomly-assigned ideas, either written by experts or generated by an LLM. Each expert spent over 100 hours implementing the idea and wrote a 4-page short paper to document the experiments. All the executed projects are then reviewed blindly by expert NLP researchers. Comparing the review scores of the same ideas before and after execution, the scores of the LLM-generated ideas decrease significantly more than expert-written ideas on all evaluation metrics (novelty, excitement, effectiveness, and overall; p < 0.05), closing the gap between LLM and human ideas observed at the ideation stage. When comparing the aggregated review scores from the execution study, we even observe that for many metrics there is a flip in rankings where human ideas score higher than LLM ideas. This ideation-execution gap highlights the limitations of current LLMs in generating truly effective research ideas and the challenge of evaluating research ideas in the absence of execution outcomes.

Financial reports offer critical insights into a company's operations, yet their extensive length typically spanning 30 40 pages poses challenges for swift decision making in dynamic markets. To address this, we leveraged finetuned Large Language Models (LLMs) to distill key indicators and operational metrics from these reports basis questions from the user. We devised a method to locate critical data, and leverage the FinQA dataset to fine-tune both Llama-2 7B and T5 models for customized question answering. We achieved results comparable to baseline on the final numerical answer, a competitive accuracy in numerical reasoning and calculation.

Personalized treatment effect estimates are often of interest in high-stakes applications -- thus, before deploying a model estimating such effects in practice, one needs to be sure that the best candidate from the ever-growing machine learning toolbox for this task was chosen. Unfortunately, due to the absence of counterfactual information in practice, it is usually not possible to rely on standard validation metrics for doing so, leading to a well-known model selection dilemma in the treatment effect estimation literature. While some solutions have recently been investigated, systematic understanding of the strengths and weaknesses of different model selection criteria is still lacking. In this paper, instead of attempting to declare a global `winner', we therefore empirically investigate success- and failure modes of different selection criteria. We highlight that there is a complex interplay between selection strategies, candidate estimators and the data used for comparing them, and provide interesting insights into the relative (dis)advantages of different criteria alongside desiderata for the design of further illuminating empirical studies in this context.

Diffusion-based Large Language Models (D-LLMs) represent a promising frontier in generative AI, offering fully parallel token generation that can lead to significant throughput advantages and superior GPU utilization over the traditional autoregressive paradigm. However, this parallelism is constrained by the requirement of a fixed-size response length prior to generation. This architectural limitation imposes a severe trade-off: oversized response length results in computational waste on semantically meaningless padding tokens, while undersized response length causes output truncation requiring costly re-computations that introduce unpredictable latency spikes. To tackle this issue, we propose Predict-then-Diffuse, a simple and model-agnostic framework that enables compute-budgeted inference per input query by first estimating the response length and then using it to run inference with D-LLM. At its core lies an Adaptive Response Length Predictor (AdaRLP), which estimates the optimal response length given an input query. As a measure against under-estimating the response length and re-running inference with a higher value, we introduce a data-driven safety mechanism based on a small increase of the predicted length. As a whole, our framework avoids wasting computation on padding tokens, at the same time preserving output quality. Experimental validation on multiple datasets demonstrates that Predict-then-Diffuse significantly reduces computational costs (FLOP) compared to the default D-LLM inference mechanism, while being robust to skewed data distributions.

The BITEM team entered both subtasks of the NTCIR-19 R2C2 task with a single agentic pipeline, in which a model searches, reads and records evidence over a movie corpus while an orchestrator holds the record and rules on what may be submitted. A claim is admitted only once an entailment cascade has checked it against the passage it cites, and an answer is released only once enough checked evidence stands behind it. Each question is run three or four times, every pass retrieving from a corpus stripped of what the earlier passes have already seen. The confidence filed with each answer is computed by the orchestrator from what the run leaves behind and is never asked of the model, which is offered no way to rate itself. The two retrieval runs placed 4th and 5th of 22, pooling the passes was worth 0.0709 nDCG@20, and the gain was largest on the multi-hop and post-processing-heavy questions, where the organisers rank the pooled run top of the field. Sixteen of the 25 answer runs were built on passages these two runs supplied, 12 of them filed by other teams. HMR rewards a system whose confidence is high where it answers right and low where it answers wrong. The pipeline reached an accuracy of 0.9219, 6th of 25, while the confidence filed with those answers gave an HMR of 0.4915, 13th. A few rules crafted over those same recorded signals, with no further model call and no further retrieval, raise that to an accuracy of 0.9375, 5th, and an HMR of 0.6985, 9th. Ranking on HMR alone can reward a system for answering wrongly with low confidence, so we propose accHMR, the accuracy multiplied by HMR, which reports the reward in proportion to the accuracy, and on which the revised rules would have scored 0.6549, 5th. For future work, fitting a model on the numbers the pipeline already produces, rather than writing such rules by hand, would be a real step forward.

The validity screen (Cacioli, 2026d, 2026e) classifies LLM confidence signals as Valid, Indeterminate, or Invalid. We test whether these classifications predict selective prediction performance. Twenty frontier LLMs from seven families were evaluated on 524 items across six cognitive tracks. Valid models show mean Type 2 AUROC = .624 (SD = .048). Invalid models show mean AUROC = .357 (SD = .231). Cohen's d = 2.81, p = .002. The tiers order monotonically: Invalid (.357) < Indeterminate (.554) < Valid (.624). Split-half cross-validation yields median d = 1.77, P(d > 0) = 1.0 across 1,000 splits. The three-tier classification accounts for 47% of the variance in AUROC. DeepSeek-R1 drops from 85.3% accuracy at full coverage to 11.3% at 10% coverage. The screen predicts the criterion. For selective prediction, the screen matters.

We introduce Instruct-SkillMix, an automated approach for creating diverse, high quality SFT data. The Instruct-SkillMix pipeline involves two stages, each leveraging an existing powerful LLM: (1) Skill extraction: uses the LLM to extract core "skills" for instruction-following, either from existing datasets, or by directly prompting the model; (2) Data generation: uses the powerful LLM to generate (instruction, response) data that exhibit a randomly chosen pair of these skills. Here, the use of random skill combinations promotes diversity and difficulty. Vanilla SFT (i.e., no PPO, DPO, or RL methods) on data generated from Instruct-SkillMix leads to strong gains on instruction following benchmarks such as AlpacaEval 2.0, MT-Bench, and WildBench. With just 4K examples, LLaMA-3-8B-Base achieves 42.76% length-controlled win rate on AlpacaEval 2.0. To our knowledge, this achieves state-of-the-art performance among all models that have only undergone SFT (no RL methods) and competes with proprietary models such as Claude 3 Opus and LLaMA-3.1-405B-Instruct. Ablation studies also suggest plausible reasons for why creating open instruction-tuning datasets via naive crowd-sourcing has proved difficult. Introducing low quality answers ("shirkers") in 20% of Instruct-SkillMix examples causes performance to plummet, sometimes catastrophically. The Instruct-SkillMix pipeline is flexible and is adaptable to other settings.

Large language models often show users a final response and a short reasoning summary while the full reasoning trace stays hidden. We introduce an observability ladder that holds each completed run fixed and varies only what a reader inspects to judge whether the answer is correct: the response, a self-summary the model writes from the trace, the trace itself, and internal signals, each with and without the prompt. Across three benchmarks and five open-weight Qwen3 and gpt-oss models, we train matched linear correctness predictors on each access level. Without the prompt, summaries carry most of the trace's ranking signal (mean AUROC 0.774 versus 0.813) and add +0.156 over the response alone. With the prompt visible, the summary's gain collapses to +0.019, while the trace still adds +0.041. Even at equal length, the trace's last words predict correctness as well as summaries, or slightly better, and carry denser and more discriminative uncertainty and self-correction cues. On MMLU-Pro questions with both correct and incorrect runs, linear summary readers are near chance and trace readers retain only modest signal, both with and without the prompt (prompt-withheld AUROC 0.503-0.545 versus 0.544-0.590). With the prompt withheld, a GPT-5-mini reader recovers substantially more signal from both summaries and traces on gpt-oss-20b, and even then the trace keeps a small +0.034 advantage. Much of the linear readers' trace signal is associated with length. In the common case where users already hold the prompt, summaries are less helpful than the full trace for monitoring correctness. Monitorability is thus a joint property of the display and the reader, so any monitorability claim, including for faithfulness, should specify both.

Large Language Models (LLMs) have shown promising performance in summary evaluation tasks, yet they face challenges such as high computational costs and the Lost-in-the-Middle problem where important information in the middle of long documents is often overlooked. To address these issues, this paper introduces a novel approach, Extract-then-Evaluate, which involves extracting key sentences from a long source document and then evaluating the summary by prompting LLMs. The results reveal that the proposed method not only significantly reduces evaluation costs but also exhibits a higher correlation with human evaluations. Furthermore, we provide practical recommendations for optimal document length and sentence extraction methods, contributing to the development of cost-effective yet more accurate methods for LLM-based text generation evaluation.

Training of a tokenizer plays an important role in the performance of deep learning models. This research aims to understand the performance of tokenizers in five state-of-the-art (SOTA) large language models (LLMs) in the Assamese language of India. The research is important to understand the multi-lingual support for a low-resourced language such as Assamese. Our research reveals that the tokenizer of SUTRA from Two AI performs the best with an average Normalized Sequence Length (NSL) value of 0.45, closely followed by the tokenizer of GPT-4o from Open AI with an average NSL value of 0.54, followed by Gemma 2, Meta Llama 3.1, and Mistral Large Instruct 2407 with an average NSL value of 0.82, 1.4, and 1.48 respectively.

We study continued training and supervised fine-tuning (SFT) of a language model (LM) to make effective use of long-context information. We first establish a reliable evaluation protocol to guide model development -- Instead of perplexity or simple needle-in-a-haystack (NIAH) tests, we use a broad set of long-context tasks, and we evaluate models after SFT with instruction data as this better reveals long-context abilities. Supported by our robust evaluations, we run thorough experiments to decide the data mix for continued pre-training, the instruction tuning dataset, and many other design choices. We find that (1) code repositories and books are excellent sources of long data, but it is crucial to combine them with high-quality short data; (2) training with a sequence length beyond the evaluation length boosts long-context performance; (3) for SFT, using only short instruction datasets yields strong performance on long-context tasks. Our final model, ProLong-8B, which is initialized from Llama-3 and trained on 40B tokens, demonstrates state-of-the-art long-context performance among similarly sized models at a length of 128K. ProLong outperforms Llama-3.18B-Instruct on the majority of long-context tasks despite having seen only 5% as many tokens during long-context training. Additionally, ProLong can effectively process up to 512K tokens, one of the longest context windows of publicly available LMs.

State-of-the-art large language models (LLMs) are now claiming remarkable supported context lengths of 256k or even more. In contrast, the average context lengths of mainstream benchmarks are insufficient (5k-21k), and they suffer from potential knowledge leakage and inaccurate metrics, resulting in biased evaluation. This paper introduces LV-Eval, a challenging long-context benchmark with five length levels (16k, 32k, 64k, 128k, and 256k) reaching up to 256k words. LV-Eval features two main tasks, single-hop QA and multi-hop QA, comprising 11 bilingual datasets. The design of LV-Eval has incorporated three key techniques, namely confusing facts insertion, keyword and phrase replacement, and keyword-recall-based metric design. The advantages of LV-Eval include controllable evaluation across different context lengths, challenging test instances with confusing facts, mitigated knowledge leakage, and more objective evaluations. We evaluate 10 LLMs on LV-Eval and conduct ablation studies on the techniques used in LV-Eval construction. The results reveal that: (i) Commercial LLMs generally outperform open-source LLMs when evaluated within length levels shorter than their claimed context length. However, their overall performance is surpassed by open-source LLMs with longer context lengths. (ii) Extremely long-context LLMs, such as Yi-6B-200k, exhibit a relatively gentle degradation of performance, but their absolute performances may not necessarily be higher than those of LLMs with shorter context lengths. (iii) LLMs' performances can significantly degrade in the presence of confusing information, especially in the pressure test of "needle in a haystack". (iv) Issues related to knowledge leakage and inaccurate metrics introduce bias in evaluation, and these concerns are alleviated in LV-Eval. All datasets and evaluation codes are released at: https://github.com/infinigence/LVEval.

When a long-horizon agent execution fails, outcome-level evaluation reveals the unsuccessful result but not where the decisive error entered the trajectory. Developers must then inspect the full execution to identify the responsible role and localize the earliest decisive root-cause step. Existing failure-attribution benchmarks largely focus on shorter traces, leaving diagnosis across hundreds of recorded steps underexplored. We introduce LongRCA Bench, comprising 1,140 failed trajectories across five domains without injected errors. It provides independently scored human labels for the responsible role and earliest decisive root-cause step. The median trajectory contains 145 steps, and the strongest baseline reaches only 13.2% exact root-step accuracy. We further present Root-Cause Trajectory Attribution (RCTA), a training-free method that retrieves candidate error steps from segment summaries and traces them to available earlier handoff instructions. Using the same backbone, benchmark instances, and scoring protocol, RCTA reaches 51.1% responsible-role accuracy and 24.1% exact root-step accuracy. These results highlight the need to evaluate responsible-role attribution and exact root-step localization as separate targets in long-trajectory failure diagnosis.

Knowing domain rules does not guarantee applying them to a case. We study this distinction in traditional Chinese Bazi through 3,000 Chinese multiple-choice questions spanning 14 Theory and 11 Case categories. Six endpoint systems are evaluated, with primary results reported on a 2,492-item model-informed refinement. Theory accuracy exceeds Case accuracy for every system, and gaps of 16.60-29.56 percentage points remain when invalid responses are excluded. The contrast is more specific than a general case-reasoning deficit. Across six systems, Twelve Stages and Nayin reach mean accuracies of 89.10% and 88.62%, while Shensha Basics reaches 75.96%. Within Case, Luck Pillars averages 84.62%, but Career and Family Relations average only 36.98% and 38.19%. Overall rankings also conceal different category strengths. On the original 3,000 items, paired DeepSeek native/disabled comparisons associate native configurations with Theory gains of 6.53 and 12.20 points for Flash and Pro, respectively; Case changes are -3.67 and +1.27 points. These are provider-configuration associations, not isolated causal effects of reasoning. The results motivate task-specific evaluation of cultural-domain applications rather than reliance on aggregate knowledge scores. The benchmark measures agreement with a model-generated, model-verified answer key, not real-world predictive validity. Final-set results are post-selection descriptions, and incomplete provenance and expert validation constrain their interpretation.

The advent of Large Language Models (LLMs) represents a notable breakthrough in Natural Language Processing (NLP), contributing to substantial progress in both text comprehension and generation. However, amidst these advancements, it is noteworthy that LLMs often face a limitation in terms of context length extrapolation. Understanding and extending the context length for LLMs is crucial in enhancing their performance across various NLP applications. In this survey paper, we delve into the multifaceted aspects of exploring why it is essential, and the potential transformations that superior techniques could bring to NLP applications. We study the inherent challenges associated with extending context length and present an organized overview of the existing strategies employed by researchers. Additionally, we discuss the intricacies of evaluating context extension techniques and highlight the open challenges that researchers face in this domain. Furthermore, we explore whether there is a consensus within the research community regarding evaluation standards and identify areas where further agreement is needed. This comprehensive survey aims to serve as a valuable resource for researchers, guiding them through the nuances of context length extension techniques and fostering discussions on future advancements in this evolving field.

Large language models (LLMs) struggle with precise length control, particularly in zero-shot settings. We conduct a comprehensive study evaluating LLMs' length control capabilities across multiple measures and propose practical methods to improve controllability. Our experiments with LLaMA 3 reveal stark differences in length adherence across measures and highlight inherent biases of the model. To address these challenges, we introduce a set of methods: length approximation, target adjustment, sample filtering, and automated revisions. By combining these methods, we demonstrate substantial improvements in length compliance while maintaining or enhancing summary quality, providing highly effective zero-shot strategies for precise length control without the need for model fine-tuning or architectural changes. With our work, we not only advance our understanding of LLM behavior in controlled text generation but also pave the way for more reliable and adaptable summarization systems in real-world applications.