new

Get trending papers in your email inbox!

Subscribe

Daily Papers

byAK and the research community

Nov 19

Latent Sparse Attention: Towards Subquadratic Ultra-Long Context Attention

Agentic workflows are increasingly shifting towards long-horizon tasks, such as open-ended scientific work, where models benefit from long context windows. However, at ultra-long context (above 1M tokens), full attention is prohibitively expensive. A promising approach is sparse attention, in which a dedicated indexer selects a subset of keys for the attention. Yet current methods of indexing either still have quadratic cost in context size or lose accuracy at ultra-long context scale. We address this with Latent Sparse Attention (LSA), an efficient attention method that maintains accuracy at ultra-long context. In LSA, keys are hierarchically clustered in a learned latent space, and queries score the hierarchy of clusters instead of scoring every key, making prefill, decode and training faster at long context. We propose two variants, Distilled-LSA (D-LSA), which can be applied to pre-trained models via distillation, and Native-LSA (N-LSA), which is trained natively on the next-token loss alone. On GLM-5.3-FLASH (QWEN-3.8-27B) at 10M tokens, D-LSA speeds up one attention layer by 22.7x (72.4x) in prefill and 4.6x (42.5x) in decode over its original sparse attention (full attention), while keeping similar accuracy across a variety of tasks. N-LSA matches the accuracy of full attention with a training speedup of 1.6x per layer at 128k tokens. LSA offers practical and efficient attention for models to benefit from ultra-long context.

  • 10 authors
·
Oct 29

How Fragile Is Safety Alignment at Frontier Scale? A Single-Direction Attack on a 320B MoE

Directional ablation removes an aligned language model's ability to refuse by projecting a single "refusal direction" out of the weights that write the residual stream. It needs no gradient-based training and no optimization, only a few hundred contrastive prompts, which makes it the canonical white-box attack on open-weight alignment. However, it has been established only on dense models up to roughly 70B parameters. We study whether it survives the shift to frontier mixture-of-experts (MoE) models whose residual streams are no longer a single tensor and whose weights ship quantized. We apply it to GLM-5.3-Flash (320B parameters, 288 routed experts, a four-wide hyper-connection residual, block-FP8). The attack survives the architecture, but what it reaches is no longer where a reader of the original recipe would look for it. Editing the attention, dense and routed-expert writers on their own removes 0.039, 0.016 and 0.148 of refusal respectively; editing all three together removes 0.776. As a result, 74% of the effect exists only under the joint intervention. The part the conventional recipe reaches by module-name matching accounts for 0.066 of that 0.776, which is why it fails silently on an MoE. The effect does not follow from removing just any direction: ablating a random direction orthogonal to it leaves refusal unchanged. A category-concentrated residue survives every edit we tried: subspaces fitted on violence, sexual content and hate leave measurable refusal at every rank from 1 to 12. We report the method, the 41-89 percentage-point reductions it achieves across seven harmful benchmarks with no detected change in capability, and the boundary where it stops.

  • 3 authors
·
Sep 8

QDA: Closed-Loop LLM Quantization Design with Agents

Post-training quantization is the standard way to serve large language models at low cost. However, the best quantization recipe differs across models and precisions. As a result, experts must redesign it for every new model, which is time-consuming and often sub-optimal. Fortunately, an LLM agent can take over this design by refining the recipe over rounds of building and evaluating candidates, so that more rounds lead to better recipes. However, evaluating one quantized reasoning model takes tens of GPU-hours, and combining methods from different codebases means porting each to the new architecture by hand. We present QDA, a system in which an agent designs the recipe for each model and precision. To make evaluation fast, QDA scores each candidate in minutes by its output KL divergence from the original model on reasoning traces that the agent never sees, and checks promising ones with long generations to catch failures such as repetition loops. To make building fast, QDA provides a library of public PTQ methods that are compatible with one another and with new model architectures. Across four reasoning models, QDA improves accuracy over the strongest public recipe by 0.9, 1.4 and 2.2 percentage points on average at W4A16, W4A8 and W3A16, using 64 GPU-hours per search. On the 320B-parameter MoE model GLM-5.3-Flash at NVFP4 W4A4, it recovers 98.3% of the FP8 model's accuracy, against 96.9% for the best vendor checkpoint. Every QDA checkpoint also runs on existing serving kernels with no inference overhead. We will release the library, the evaluator and the harness.

mit-han-lab MIT HAN Lab
·
Nov 16 1

Intent Contracts: Taking Tools Away from Agents

Intent contracts are an authorization interface for tool-using LLM agents in which the planning model holds no tools. For each step it submits a short logged contract: a goal, a verbatim quotation of the user, and scope fields. Code validates the contract, and a small trusted model checks it against the user's words and writes the tool call itself. Provided the validation and dispatch code is correct, every executed call is written by the trusted model, the planner's influence on a call is confined to a contract logged before it, and the trusted side decides how much of a request to carry out. We compare contracts with two reviewers of planner-written commands: one sees only the user's words and the command, the information boundary of a reasoning-blind permission classifier, and one also sees the planner's goal and a checked quotation. With gpt-6.1-terra as planner and GLM-5.3-Flash as the trusted model on three AmPermBench task families, every object changed under contracts is named in the goal of the contract that produced the change; under review the planner names objects in its commands and rarely in its goals. When the planner adds work the user never asked for, contracts complete the user's task in 27/30 runs and the intent-visible reviewer in 16/30, because a reviewer can only refuse a whole command. Contracts do not reduce unauthorized actions. They are indistinguishable from blind review, and once the benchmark's environment terms are clarified, every condition that changes state hits a wrong target in 40-49% of those runs. That error is the planner's choice of objects for an ambiguous reference, which no trusted model here can see the environment to check; telling the executor to decline when unsure does not change it. We evaluate benign over-reach with one pair of models, not adversarial planners. We release the plugin and run records.

  • 4 authors
·
Nov 17

Continual-ARC: Measuring Skill Acquisition over a Working Lifetime

Agent harnesses have emerged as a promising mechanism for continual learning, alongside test-time parametric learning. However, the current benchmarks have several limitations: they generally focus on a single applied domain, serve each instance only once, and assume a single family of learner, so memory-based and weight-based systems are rarely compared on the same scale. We propose Continual-ARC, a lifelong stream built from the three ARC-AGI editions, whose tasks need no domain knowledge. The same tasks return after dormant gaps with fresh instances, while new tasks keep arriving. On each instance, the learner may pay for demonstrations, submit answers, or act inside the harness, with retries permitted on incorrect submissions, which are paid for. Each instance is scored based on the share of this feedback budget consumed. Since a learner may keep anything between instances, memory-based and weight-based systems are scored on the same scale. After the stream, the learner is frozen and tested on fresh instances of the tasks it met, which measures memorization, and on unseen tasks, which measures generalization. We run an agent harness with frozen weights, with and without persistence, on Qwen3.5-9B and GLM-5.3 Flash, comparing them with two prompting floors, a program-induction method and Qwen3.5-9B's weights trained to imitate GLM-5.3 Flash. Persistence in the harness generally does not improve memorization unless provided with explicit expert guidance. Most harness-based methods still learn less than a floor that simply caches each task's demonstrations, across all models we tested. No harness-based method gains on unseen tasks, whereas weight-based learning shows substantial gains. This opens the door to future weight-based continual learning that could teach models to drive harnesses in new domains.

unifyai Unify
·
Nov 10 6

ServeLearnBench: How Well Can Agents Self-Improve from Serving Experience?

Large language model agents are increasingly deployed to perform complex tasks in real-world environments. However, the knowledge required for correct behavior in these environments is often implicit, undisclosed, and subject to change over time. Recent continual-learning harnesses seek to address this challenge by enabling agents to improve from serving experience. Yet the effectiveness and limitations of these methods are not yet well characterized. Existing benchmarks provide only partial coverage: some explicitly provide the target knowledge, others assume a static environment, and those that support continual adaptation remain limited in scale and knowledge diversity. To enable systematic evaluation, we formalize an evolving-environment streaming dataset (EESD), in which agents must infer, apply, and revise latent environment knowledge from interaction and outcome feedback as hidden policies evolve, and introduce ServeLearnBench, spanning retail support, banking, and sales-pitch generation with 53 environment windows and 7,718 tasks. We evaluate five learning harnesses (RAG, Mem0, SkillOpt, Continual Harness, and Prime) across six models (GPT-5.6 Terra, Opus 5, Kimi K3, GLM-5.3, DeepSeek V4.1 Flash, and GLM-5.3 Flash), covering 28 model-harness pairs and 252 learning runs. Our evaluation reveals three main findings: a substantial gap remains between task capability and learning from experience; continual adaptation is costly and can degrade already-correct behavior; and insufficient exploration emerges as a key bottleneck to effective adaptation. Overall, ServeLearnBench provides a controlled testbed for diagnosing these limitations and tracking progress toward agents that continually and reliably improve through serving experience.

  • 5 authors
·
Oct 5

FrugalEvo: Towards Cost-Aware LLM-Guided Program Evolution

LLM-guided evolutionary methods, such as AlphaEvolve, have emerged as powerful approaches for challenging computational optimization problems, such as circle packing. However, prior work typically optimizes performance gain over a fixed number of iterations. We argue that practical optimization should maximize gain per unit cost. To this end, we propose FrugalEvo, a cost-aware evolutionary framework where a stronger, higher-cost LLM explores solution strategies, and a cheaper LLM implements them and iteratively refines the resulting code. We also design a cache-efficient evolution process, where our harness and prompts maximize the sharing of prefixes across different evolution steps, to improve cache reuse. To measure solution quality throughout a fixed cost budget, we introduce Budget-Aware Area Under the Curve (BA-AUC), defined as the area under the best-so-far evaluation score curve over cumulative LLM cost, up to the budget. Across 10 mathematical and systems optimization tasks, FrugalEvo matches or surpasses state-of-the-art baselines, including OpenEvolve, ShinkaEvolve, AdaEvolve, and EvoX, in final solution quality and achieves higher BA-AUC on 9 tasks. It also achieves higher average performance than these baselines on 10 algorithmic optimization tasks from ALE-Bench-Lite. Notably, on circle packing, FrugalEvo achieves new state-of-the-art performance with GPT-5.6 Terra and Luna for only 1.68 USD and with GLM-5.3 and its Flash variant for only 0.55 USD, matching or surpassing all baselines, including multi-agent methods such as CORAL and SwarmResearch, which cost approximately 50 USD on average.

  • 8 authors
·
Oct 1 2

When the KV Cache Is Slower Than the GPU: Scaling Remote KV Tiers for Context-Parallel Prefill

Reusing the attention KV cache across requests is a standard way to cut prefill cost in long-context LLM serving, and production systems increasingly back the per-engine cache with a remote tier shared by every engine. We study such a tier in a disaggregated deployment of GLM-5.3, an FP8 mixture-of-experts model with multi-head latent attention (MLA), on 8xB300 nodes running prefill context parallelism (PCP8). Enabling the remote tier (LMCache with a Valkey store) left median time-to-first-token unchanged but raised the 16k-token p99 from 3.96 s to 17.3 s. Replaying a production request stream on an isolated engine, we find that a remote hit took 0.88 s against 0.31 s to recompute the prompt, and that synchronous loading turned a hit-heavy stream into an unstable queue: p99 83.32 s with the GPUs 7% busy. A two-term cost model explains both: a hit pays off only if the tier delivers more than B* = sR (2.53 GB/s for this model and GPU), and a fixed per-hit engine cost (T0 ~ 0.33 s) means that hits on prompts shorter than about 17k tokens cannot beat recompute at any bandwidth. Guided by the model, we (i) shard the MLA KV store across PCP ranks, with a proof that the existing minimum-over-ranks lookup returns the longest cached prefix and a load protocol that can never desynchronize collectives; (ii) partition the remote tier by rank, one server instance per rank with no coordination, which raises GET bandwidth 6.3x to 10.76 GB/s; (iii) add asynchronous loading to the sharded store; and (iv) fix three client bugs that caused unbounded stalls. On the hit-heavy stream, with 98.5% of prompt tokens served from the remote tier, p99 falls from 83.32 s to 0.67 s. The sharded store raises prefill throughput by 14%, and by 34% together with an upstream kernel fix. All changes are submitted upstream to LMCache.

CoEvo-RSI: When Harnesses Evolve, Which Models Should Follow?

Model--harness co-evolution has emerged as a prominent paradigm for recursive self-improving (RSI) agents, jointly evolving their execution scaffolds and intrinsic capabilities. We study the persistent yet often overlooked brain--body mismatch problem in this paradigm, where an evolving harness (body) demands behaviors its model (brain) has not learned or restricts capabilities it already possesses. We introduce CoEvo-RSI, a RSI framework that advances model--harness co-evolution through agent-driven, bidirectional historical inheritance. Drawing on a shared evolutionary archive and execution feedback, a meta-agent coordinates harness development from historical designs with model checkpoint selection or merging and subsequent adaptation under the selected harness, aligning inherited capabilities with evolving execution requirements to reduce brain--body mismatch. Using only supervised fine-tuning (SFT) on its own rejection-sampled trajectories, CoEvo-RSI makes Qwen3.5-9B competitive with larger models across diverse vertical applications, including multi-agent failure attribution, role-aware text-to-SQL, and smart-home instruction following. It surpasses DeepSeek-V4-Pro on TraceElephant (+14.16 agent accuracy) and GLM-5.3 on HomeBench VM (+4.90 success rate), underscoring the need for brain--body alignment and the potential of bidirectional historical inheritance to advance agent co-evolution.

  • 8 authors
·
Oct 17 3

Safe Actions Alone Do Not Ensure Safe Agents: Identifying Unfulfilled Obligations with Guard Models

Guard models are increasingly used to safeguard LLM-based agents, primarily by identifying actions that agents are forbidden to perform. However, identifying forbidden actions alone is insufficient to ensure agent safety. In this paper, we argue that agent safety also depends on identifying required yet unperformed safety-critical actions, which we call obligations. Our preliminary study on a popular benchmark for evaluating safety shows that 56.92% of GLM-5.3 trajectories contain unfulfilled obligations, compared with only 30.00% containing forbidden actions. This finding reveals unfulfilled obligations as a major and previously overlooked source of safety risk. However, to our knowledge, no existing benchmark evaluates whether guard models can identify these obligations. To close this gap, we introduce ObligationBench, the first benchmark for evaluating the capability of obligation identification, comprising 240 expert-validated trajectories covering issue resolution, feature development, and terminal operations. Our evaluation of 14 representative models reveals substantial limitations: the highest recall and exact-match rate are only 48.97% and 10.00%, respectively. To address these limitations, we develop ObligationGuard using 40,000 synthetic training examples. ObligationGuard achieves 57.52% recall and an exact-match rate of 21.67%, surpassing all evaluated models on both metrics. We call on the community to incorporate obligation identification into the design and evaluation of future guard models to improve agent safety.

  • 4 authors
·
Oct 7

DecisionPivot: A Diagnostic Benchmark for Decision Models in LLM Systems

Decision models support routing, guardrails, and output verification in LLM systems. However, item accuracy alone cannot establish whether models make correct decisions across controlled variations or assign reliable confidence to their predictions. We introduce DecisionPivot, a diagnostic benchmark comprising 48 tasks and 40,033 items across seven decision roles. Its core track contains 7,800 items from 13 generated task families, organized into contrast groups with computed answers verified by independently implemented solvers. The complementary reference track contains 32,233 items from 35 public test sets, preserving their original evaluation metrics. We evaluate 40 decision models and 14 general models using documented, model-specific protocols, assessing item and contrast-group accuracy, probability ordering, calibration, and empirical selective coverage. Our results reveal substantial differences obscured by aggregate scores: two 27B models separated by only 2.6 points on the reference track achieve contrast-group accuracies of 70.1% and 44.2%. Across 2,400 binary contrast pairs, Perplexity Decider correctly ranks the positive member above the negative member in 92.3% of pairs but answers both members correctly in only 74.1%, exposing a gap between probability ordering and decision correctness. Models with similar item accuracy also differ in the reliability of their high-confidence predictions. On a separate 1,296-item subset, GLM-5.3 achieves 93.5\% contrast-group accuracy at a higher reasoning setting, compared with 81.5% at its lowest effort. By linking errors to pivotal details and difficulty levels, DecisionPivot provides fine-grained diagnostics of decision failures that aggregate evaluation alone cannot reveal.

startlux-models StartLux
·
Oct 19 3

CoworkOS:Scaling Agent Coworkers for Real-world Automation

Rapid advances in cowork agents are shifting human-AI collaboration toward real-world automation, delivering software, designs, and other artifacts. However, such productivity work demands both high-quality delivery on complex productivity tasks and continued monitoring and decision-making over long horizons, a combination that remains an open challenge. We introduce CoworkOS, a general-purpose harness for long-horizon work that separates task strategy from execution management: the model decides actions, delegation, and acceptance, while the runtime maintains state and feedback. Its three-layer architecture enables direct execution or delegation to Subagents, synchronously or asynchronously, under stable responsibilities. Through a shared application layer, humans and agents inspect and revise the same evolving documents, presentations, CAD models, and Blender scenes, so human-in-the-loop participation can enter both through conversation and through direct modification of the work itself. To sustain continuity across agents and sessions, CoworkOS further introduces hierarchical workspaces and persistent semantic memory, so working files, key decisions, and long-running Job logs remain shareable, reusable, and resumable after interruption. Across 12 benchmarks and 6 categories of real-world applications, CoworkOS proves highly competitive with frontier agent applications such as Codex and DeepSeek Harness. As a sustained-work case study, CoworkOS uses GLM-5.3 as its backbone to conduct RSI-driven automated post-training on Qwen3.6-35B-A3B. Over five rounds, the model's average score across eight benchmarks rose from 33.19 to 42.95 (+9.76). Looking ahead, we believe CoworkOS opens a meaningful path toward general cowork-agent intelligence.

CRJudgeBench: Can AI Detect Plausible but Invalid Code Reviews?

Large language models can generate plausible code-review comments, but such comments may contain technically incorrect claims that mislead developers. We study technical trustworthiness judgment: determining whether a review comment's core technical claims are correct and applicable to the reviewed code in its repository context. Existing code-review benchmarks primarily evaluate review generation, issue discovery, or general comment quality, but do not directly assess whether an agent can determine the technical trustworthy of an individual review comment. To fill this gap, we introduce CRJudgeBench, a benchmark of 1199 instances constructed from real pull requests and expert-verified perturbations, covering both trustworthy and plausible but untrustworthy comments. We further present Sentinel, a repository-grounded agentic judge that actively gathers code evidence to verify review comments before making judgments. Starting from Qwen3-Coder-30B-A3B-Instruct, Sentinel is trained on the CRJudgeBench training split through iterative action-level learning from a privileged teacher. On the 359-instance CRJudgeBench test set, Sentinel achieves 76.60\% accuracy, outperforming GLM-5.3 by 6.13 percentage points and its base model by 19.78 points. These results show that even state-of-the-art general-purpose LLMs struggle to identify untrustworthy comments, while iterative action-level learning substantially improves the accuracy of repository-grounded trustworthiness judgments. Our dataset is available at https://huggingface.co/datasets/dcloud347/CRJudgeBenchmark

  • 5 authors
·
Sep 28

Priced Guidance: Can Language Models Generate Future Research Ideas?

We evaluate language models' capability to generate novel research ideas through the lens of compression. We aim to lower-bound the potentially tiny probability that a language model generates the essence of a future research idea without any hints. Rather than estimate this probability through expensive repeated sampling, our Priced Guidance framework measures the compression cost: how many additional bits of information are needed to guide the model to recover the target idea. We prove that if the model can recover the target idea with at most K bits of guidance in expectation, then it can generate the idea without any guidance with probability at least 2^{-K}. In our framework, the language model, called the generator, can pose a sequence of multiple-choice questions and specify a probability distribution over possible answers. A guide, which is a language model with access to the target idea, selects answers. If a selected answer has prior probability p, the generator pays -log_2 p bits. The generator aims to produce an idea that matches the essence of the target idea with minimal cumulative cost. This cumulative cost equals, up to an additive constant, the number of bits of information sent by the guide. Using this methodology, we evaluate five generator LLMs (Opus 5, Fable 5.1, GPT-6 Astra, GPT-5.6 Sol, and GLM 5.3) on the core ideas in 87 recent high-quality deep learning papers and we use LLM as judge to determine whether the generated idea matches the target in terms of the central research object and defining mechanism. Fable 5.1 achieves the lowest median compression cost at 69.9 bits, substantially lower than gzip's median of 5,712 bits for losslessly compressing the summary of the target idea. A uniform ensemble of Fable 5.1, Opus 5, and Astra further reduces the median compression cost to 55.8 bits and improves the generation probability lower bound by 18,000 times.

  • 3 authors
·
Oct 3

Big-Big-SycoBench: A Benchmark to Detect Sycophancy Through Model Responses to Valid vs. Invalid Arguments

Sycophancy in large language models risks amplifying unsupported claims or beliefs across the knowledge ecosystem as these systems increasingly contribute to research synthesis, evaluation, and decision-making. A reliable model should revise its judgement when a user identifies a genuine flaw, but should resist pushback based on fallacies, emotion, or irrelevant appeals to authority. While sycophancy is usually measured on questions with a verifiable answer, the evaluative tasks in which it is most difficult to detect have no such ground truth. We introduce Big-Big-SycoBench, which measures sycophancy through argument discernment: the ability to revise an evaluation after valid criticism while resisting invalid pressure. The benchmark consists of 450 text artefacts with 5,400 arguments for raising or lowering their quality scores, each reviewed by one or more of twelve human annotators. An LLM scores each artefact twenty times to establish its baseline score and uncertainty. The model is then given a valid or invalid argument and asked to update its score. We only count a revision if it exceeds the model's own baseline variability. We use the resulting score delta to distinguish responsiveness to valid arguments from compliance with invalid ones. Across ten models, responsiveness is 88-95\% for seven, 76-77\% for Kimi K3 and GLM-5.3, and 32\% for Opus 5.5. Compliance ranges from 0.2\% for GPT-6 Astra to 95\% for GPT-4o. Big-Big-SycoBench is intended to guide the development of models whose judgements can be corrected by sound arguments, but not displaced by unsupported pressure.

  • 14 authors
·
Oct 10