paper_id stringlengths 10 10 | items listlengths 4 82 | gt float64 1 9.5 |
|---|---|---|
xaafWdM5jI | [
"strength: The paper addresses limitations of a previously published method, FourierGNN.",
"strength: [S1]: The authors showcase the importance of spectral graphs for time series forecasting while highlighting the importance of sparsity in the structure.",
"strength: [S2]: The proposed framework remains simple ... | 3.5 |
fUGhVYPVRM | [
"strength: The proposed framework is flexible and simple enough to generalize to different properties and provides a generic way to makes the CTC training more controllable.",
"strength: For the low latency application, the proposed framework achieves better latency and quality tradeoff than a few existing approa... | 7 |
InRaT76E2S | [
"strength: The authors conducted extensive experiments across CV and NLP tasks to demonstrate the effectiveness of activation decay (AD).",
"strength: The paper clarity is good and the motivation to move away from stochastic regularization is important.",
"strength: S1: Timely, well suited topic for application... | 2.5 |
f89YIjbuRC | [
"strength: The paper improves the projector by taking advantage of the autoencoder's ability to capture high-quality representations, which improves the performance of the contrastive learning method.",
"strength: The experiments conducted could be a good reference for industry applications that don't want to fin... | 3.5 |
7visV100Ms | [
"strength: The strength of the paper is that it offers a sound solution in terms of introducing diversity in the self-improvement process. While this is not exactly novel, but I do see a world where this works and improves model progressively. Moreover, use of keyword means that it provides a transparency in what w... | 6.6 |
fswihJIYbd | [
"strength: This paper introduces the Adaptive Decomposition Prompt Tuning (ADePT) method, which innovatively addresses the generalization limitations caused by fixed token embedding offsets in traditional DePT methods.",
"strength: ADePT achieves excellent adaptability without increasing inference time or requiri... | 7 |
ZULjcYLWKe | [
"strength: The proposed framework uses the latest diffusion model to solve offline RL with state perturbations. I believe there is a strong motivation behind this approach. Indeed, an important RL problem is studied with state-of-the-art tool.",
"strength: The paper is well-written. The motivations behind each se... | 7 |
1ebgtm7P10 | [
"strength: The paper is straightforward: the authors identify a problem (reduced OOD detection performance models from a certain training script), provide an explanation (mixup and label smoothing) and a fix for it",
"strength: The experiments identifying label smoothing and mixup as the problems are convincing a... | 4.25 |
8SaFvd4sj2 | [
"strength: While most previous work focuses on multi-domain generalization, this paper tackles the more challenging and practical problem of single-domain generalization. This is especially relevant for autonomous driving where collecting multi-domain data is expensive and time-consuming.",
"strength: The paper p... | 4.8 |
hLIlN0f4ix | [
"strength: Novel idea of using language decoders as text feature extractor to pair with frozen vision encoders just with a small MLP projector. This paper shows that decoders outperforms encoders for visual concept representation.",
"strength: This paper shows correlation between language tasks and its vision tra... | 5 |
8HuLgtjqOD | [
"strength: Using the random seed for reconstructing the random projections is a smart trick for efficient communication and saving space.",
"strength: This method has the benefit of being an unbiased estimator of the true gradient, which is not the case with low-bit and some low-rank methods. Even though the vari... | 6 |
G5DziesYxL | [
"strength: The paper is well-motivated: Most previous works has primarily focused on text datasets, or a single feature or dataset. The paper conducted a multimodal and multi-feature dataset, which addresses this gap.",
"strength: The scale of this study is impressive: It covers nearly 4000 datasets, 3 modalities... | 6.5 |
eW4yh6HKz4 | [
"strength: The paper offers an interesting insight: full model quantization introduces inter-layer correlations.",
"strength: Experimental results on LLAMA1, LLAMA2, and OPT demonstrate substantial improvements in W2 and W4 settings, alongside a significant reduction in PTQ scenario.",
"strength: I have careful... | 7.6 |
FPfCUJTsCn | [
"strength: The idea of replacing solvers in the training pipeline is intriguing. Indeed, I can envision many problem classes where off-the-shelf solvers may underperform compared to simple gradient-descent-based algorithms. The proposed method could be highly effective for such problems.",
"strength: The strength... | 7.2 |
xA25Ib7H8U | [
"strength: Studying neural networks from a more geometric perspective is a promising avenue that could lead to novel insights into the inner workings of neural networks. Leveraging the powerful tools from differential geometry such as Ricci flows and Ricci solitons is a helpful contribution that could foster more s... | 2.333333 |
1MjOlHwCE6 | [
"strength: The overall idea is interesting and offers a complimentary perspective.",
"strength: (++) **Scalability Improvement**: The proposed complexity reduction from $O(n^2)$ to $O(n \\log(n))$ is a notable improvement.",
"strength: (++) **Practical Utility**: The paper demonstrates comparable or slightly be... | 2.5 |
dapU3n7yfp | [
"strength: This paper proposes a novel method of textual adversarial attack for generation tasks, which looks generalizable for different text generation tasks.",
"strength: The empirical results look convincing, outperforming the baselines by a large margin",
"strength: This paper focuses on AI safety, which i... | 5.5 |
xw3fStKCwm | [
"strength: The paper shows that TT decomposition of high dimensional data can be constructed in a way that helps achieve a high rate of compression.",
"strength: The authors then demonstrate that the compressed representation can be useful for two applications -- (1) out of distribution sample detection or anomal... | 3.75 |
cagNCwQEEN | [
"strength: The paper is well-written and easy to understand, making it accessible to a wide audience.",
"strength: The use of State Space Models (SSMs) to reduce inference costs is a reasonable and potentially beneficial approach.",
"strength: The hybrid transformer-Mamba architecture effectively manages long c... | 3.4 |
Qyile3DctL | [
"strength: The compilation of a large dataset containing both correct and incorrect candidate solutions from multiple LLMs, covering math and coding tasks, offers a valuable resource for training solution verifiers.",
"strength: The CoTnPoT verification strategy improves accuracy by leveraging the complementary s... | 5 |
RXU6qde675 | [
"strength: The paper is well organized and easy to understand.",
"strength: The method as a framework can be generalized easily to other tasks on the multi-layer networks.",
"strength: Extensive experiments on real-world datasets show the propsed method can effectively and efficiently capture the information be... | 2.5 |
tiKJsepvr0 | [
"strength: The authors' core insight about efficient data usage in FL clients is valuable, as reducing unnecessary data sharing while maintaining performance is an important practical concern in real-world FL deployments.",
"strength: Despite limitations, the paper provides extensive experiments across multiple s... | 2.5 |
0VKEJKKLvr | [
"strength: The overall problem was stated clearly, and the introduction in Section 1 was well-written.",
"strength: The approach itself and the figure illustrations made sense and helped readers understand the material.",
"strength: Overall, the application of graph neural networks to this application area is c... | 3 |
15lk4nBXYb | [
"strength: The idea of converting camera poses into pixel-wise embeddings is novel, which allow the video generation model to effectively understand the camera motion.",
"strength: This paper studies three common ways of incorporating camera pose embedding into a DiT model, which could be useful for future work."... | 3 |
F3Migaak2i | [
"strength: This paper addresses a highly valuable problem, offering practical benefits for detecting model plagiarism and enhancing the robustness of AI model development.",
"strength: It introduces a promising analytical framework that evaluates prediction differences across the entire input space, enabling a co... | 3 |
XBF63bHDZw | [
"strength: This is an intriguing question and a useful framework for practical SiMT scenarios. While SiMT has been extensively studied in the research community, most papers have focused on automatic metrics rather than real human preferences. This work proposed 5 human preferences that are closely related to the e... | 7 |
xtzqU9FgSi | [
"strength: The paper compares multiple self-supervised learning approaches for sentence embeddings, providing insights into the effectiveness of different data augmentation strategies, especially text crops.",
"strength: It demonstrates that SSL can achieve sentence embedding quality close to supervised models, s... | 4.333333 |
moXtEmCleY | [
"strength: Clear Algorithm Characteristics. The work presents a clear and well-defined algorithm without making excessive claims. Both the algorithm description and the experimental design are straightforward and easy to follow.",
"strength: Experimental results effectively support the Authors' Claims.",
"stren... | 6 |
RuY1r1PDdQ | [
"strength: This paper is technically sound and proposes a benchmark for evaluating intent hallucination. The idea is generally valid but many parts of the paper are very hard to understand. See weakness for details.",
"strength: The motivation of the paper is reasonable. Current evaluation benchmarks may usually ... | 3 |
87YOFayjcG | [
"strength: This work tries to tackle the critical problem of evaluating open-ended answers of LLMs and curate a large-scale dataset with annotation of GPT-4.",
"strength: The proposed judge model is trained with two data augmentation tricks, i.e., answer swap and reference dropping, to be more robust to the answe... | 5.25 |
3fGtV4Zfgq | [
"strength: RBMs are an important model and finding appropriate ways to train them is a topic of significant interest.",
"strength: The paper highlights the phenomenon of critical slowing down and how pre-training the model with a low-rank approximation of the parameter matrix can help the model overcome some of t... | 4.25 |
5WEpbilssv | [
"strength: The method appears to be sound, combining RAG and COT prompting with knowledge graphs for handling complex biological relationships.",
"strength: The proposed ramework emphasizes interpretable outputs, which is beneficial to biological research.",
"strength: The evaluation was thorough, with both gra... | 4.6 |
2whSvqwemU | [
"strength: The paper makes an interesting attempt to apply Flow Matching, a technique that has shown promise in image generation, to the complex domain of time series generation.",
"strength: The paper claims substantial efficiency gains over diffusion-based methods. Diffusion models, while powerful, suffer from ... | 3 |
7gUrYE50Rb | [
"strength: The paper addresses a new and interesting problem of embodied question answering which is key to enabling human-robot interactions in their natural environments.",
"strength: Inspired by the observation that communication between humans is a combination of both verbal and non-verbal gestures (gaze and ... | 8 |
AMVLOv30Qg | [
"strength: The work proposed a reference-guided unbounded scene inpainting pipeline that leverages priors from reference-guided inpainting and depth models. The truly unseen mask-based inpainting can reduces the inpaintig inconsistency compared with object masked based approaches.",
"strength: The paper provides ... | 3.333333 |
xJ5N8qrEPl | [
"strength: The paper is well-written.",
"strength: The authors propose a single loop Hessian free method that we refer to as Lagrangian Value function-based Hessian-free Bi-level Algorithm (LV-HBA).",
"strength: The authors provide non-asymptotic convergence analysis for LV-HBA, relaxing the underlying assumpti... | 6.4 |
EWKPEtwjTy | [
"strength: the trap and cheese problem is a nice illustration of the limitation of Gaussian policies",
"strength: Whether actions should be represented with discrete or continuous variables is very compelling. Each of the two representation have their own advantages and disadvantages. Discrete actions, in contras... | 2.5 |
j9VVzueEbG | [
"strength: Originality : The paper introduces a novel use of Z-order curves for dimensionality reduction in attention mechanisms, adding a fresh approach to efficient top-k attention.",
"strength: Quality: Methodologically sound with theoretical insights like the Johnson–Lindenstrauss Lemma for dimensionality re... | 7 |
M8gXSFGkn2 | [
"strength: The paper is clear and well-motivated. The authors first diagnose current EgoVLMs' capability on the proposed EgoHOIBench benchmark and then propose a method to address the problem. This makes the paper easy to follow and provides insights to readers.",
"strength: The proposed EgoHOIBench provides a ta... | 7 |
SI6zocV2SS | [
"strength: The idea of utilizing Hebbian learning to help the model continually learn is interesting, as Hebbian learning is a well-known synaptic plasticity rule in animals. Experiments might provide insights into why these mechanisms can help animals continually learn.",
"strength: The paper connects the Hebbia... | 1.5 |
yLgr02IsXY | [
"strength: MoE is becoming very relevant for the current scaling of models as dense models are impractical at the trillion parameter scale, which makes this work relevant.",
"strength: Code is provided that is an extension of the MegatronLM code. I skimmed their AMPipe implementation and it seems fairly simple w... | 5.25 |
4qRCiEZGKd | [
"strength: The targeted problem is interesting",
"strength: The proposed approach can be easily reproduced",
"strength: This work explores the important field of neuro-symbolic reasoning, which is crucial for advancing knowledge representation and reasoning, especially for real-world applications where incomple... | 3.4 |
NY3HzOOL3u | [
"strength: Using LLMs to construct a DAG is interesting and original",
"strength: Their method seems fairly sample efficient, and it seems promising to extract domain specific knowledge from LLMs for the purpose of planning.",
"strength: Within the Minecraft game context, the Plan4MC scheme of learning fine-gra... | 4 |
QGGNvKaoIU | [
"strength: Authors provided a comprehensive related works section.",
"strength: The problem setting is clear and overall reasonable.",
"strength: The theoretical analysis is good and comprehensive, although I can't fully check the correctness.",
"strength: The empirical validation is comprehensive too.",
"s... | 7 |
KD5nJUgeW4 | [
"strength: General games are hard. Any work that attempts to learn and prove convergence in this setting is important to the community.",
"strength: Last-iterate convergence is valuable, especially compared to average-iterate algorithms that are common in the field.",
"strength: Experimental results look impres... | 7 |
fTdhM7q1o2 | [
"strength: This paper proposes the BTT model in RLHF for more reliable preference learning.",
"strength: This paper provide the theoretical results for the readers to better understand the effectiveness of the proposed method.",
"strength: This paper is well written and easy to be understanded.",
"strength: I... | 3 |
hfRb6yC0W0 | [
"strength: This paper studies an important scientific question of identifying latent signal sources associated with speech processing from non-invasive MEG data.",
"strength: The authors made improvements upon prior work to incorporate 3D spatial filter using spherical harmonics and added a temporal filter.",
"... | 3 |
FxNNiUgtfa | [
"strength: Originality: Introducing a framework to measure language model capacity in bits per parameter is a novel approach that adds a quantitative dimension to model evaluation.",
"strength: Methodology: The use of controlled synthetic datasets allows for the isolation of specific variables, providing clarity ... | 7.25 |
pzElnMrgSD | [
"strength: The research question “how to properly warp a Gaussian noise map” introduced by this paper is interesting scientifically.",
"strength: The problem is well motivated with clear explanations on why the problem is not trivial, i.e. why the conventional warping operations are not suitable. The proposed sol... | 7.5 |
S2WHlhvFGg | [
"strength: The authors address an important problem in drug repurposing and provides a strong mathematical framework for the newly proposed method. The proposed model also make use of both the protein and drug sequence to predict novel drug-target interactions. The proposed mechanisms are interesting and add over t... | 3 |
UyPmWupphV | [
"strength: Good research problem. Researchers often optimize neural architectures and hyper-parameters separately, leading to a sub-optimal solution. This paper aims to jointly optimize both, and can potentially lead to better results.",
"strength: Intuitive ideas. This paper is not the first to jointly optimize ... | 3 |
xTrAA3UKPa | [
"strength: SWGA introduces a combination of genetic algorithms with a sliding window approach tailored specifically for time series forecasting, addressing the distribution shift and non-stationarity challenges unique to this domain.",
"strength: The authors have provided detailed computational procedures by prov... | 2 |
eJhgguibXu | [
"strength: Leveraging imperfect models of the world for RL is an important problem that interests the research community.",
"strength: The proposed approach of exploring by building action priors based on successful imagined trajectories is an interesting idea.",
"strength: Most of the paper is well-written and... | 2.5 |
MJWJoICJQh | [
"strength: The proposed method merges imitation and reinforcement learning, combining both into a single objective function without requiring a separate pre-training stage.",
"strength: It selectively prioritizes high-quality policy rollout data to complement the original demonstrations, thereby providing more di... | 3.4 |
9h5paerJxC | [
"strength: The idea of designing a pipeline for the complex spatiotemporal data and high-dimensional feature spaces is intriguing.",
"strength: The description of the challenges in explainability of spatiotemporal land surface forecasting is technically sound.",
"strength: The work is original in the sense that... | 3.5 |
rWQDzq3O5c | [
"strength: Originality: This paper introduces a novel use of a linear Transformer, to perform core graph algorithms like electric flow and eigenvector decomposition without explicit positional encodings.",
"strength: Quality: The paper offers rigorous theoretical analysis with explicit weight constructions and er... | 5.75 |
Fk4Op9wpEp | [
"strength: This paper focuses on an interesting question, which as great significance to downstreaming research and tasks.",
"strength: The proposed method seems to be able to generate human images better aligned with input pose conditions and the OKS is improved compared to baseline.",
"strength: The proposed ... | 3 |
Acvo2RGSCy | [
"strength: Exposition and writing in general is in good condition. (There are small rooms for improvement in organisation and clarification.)",
"strength: The whole framework is written with careful formal details, thanks to taking decision-theoretic frame, which is a great idea to start with.",
"strength: Th... | 7.333333 |
i5Gxilzk0u | [
"strength: MOCO is quite straightforward and achieves better realism and coherence compared to existing methods that process modalities through weighted averaging.",
"strength: Comprehensive evaluation: The use of a purpose-built multi-modal benchmark and a wide range of metrics for evaluation provides a thorough... | 5.8 |
XhYWgjqCrV | [
"strength: The paper is well-written, comprehensively introducing the motivation and method details.",
"strength: The experiments are comprehensive, covering several popular vision tasks as well as varies of network scales.",
"strength: The experimental and visualized analysis is good, helping the reviewer bett... | 6.5 |
OfjIlbelrT | [
"strength: The key idea that the complexity of input data varies so we should dynamically adjust sparse attention patterns and computational budget with respect to different input examples is interesting and reasonable.",
"strength: Experimental results indicate that FlexPrefill achieves comparable results with p... | 8 |
aoW5Sm8Op8 | [
"strength: This work adresses an important and interesting problem, and proposes a thorough overview of the problem of causal inference with right-censored data, and some of the challenges associated with different types of bias. The simulation study covers several varied datasets. For all topics covered (estimands... | 2.333333 |
UB03wcP8RH | [
"strength: The paper presents a novel approach for combining multi-task and contrastive learning, which to my knowledge has not been done before. The use of multiple projection heads and learned weighting scheme specifically for handling multiple disjoint similarity metrics is creative and original.",
"strength: ... | 5.25 |
INqLJwqUmc | [
"strength: While numerous interpretability methods exist, they cannot cope well with the multimodality od data.",
"strength: The proposed methods addresses this issue, with 9% improvement for text interpratability, and 58% improvement fo image interpratability.",
"strength: The proposed method also eliminates r... | 5.25 |
fkNsgI1nye | [
"strength: A new method called SDU is introduced.",
"strength: For model comparison, it is necessary to derive the distribution $p(z|y)$ using the encrypted condition $y$. The idea here is that this can be approximated by using randomly sampled $p(z)$, which provides a similar outcome.",
"strength: The problem ... | 3 |
af2ztLTFqe | [
"strength: The paper is well-written, well-motivated.",
"strength: Extensive experiments are conducted on several aspects, with proper ablation and analysis. It shows that the methods not only improve the scores, but also generate fluent questions.",
"strength: The added cost of inference is acceptable giving t... | 6 |
Ob0UafH2YI | [
"strength: The improvement process in the methodology section is very clear. It first establishes the incapability of vanilla FedAvg to solve compositional optimization (CO) problems. Based on the identified issues, the authors propose sharing $y_k$ in each iteration and modifying the classical FedAvg aggregation t... | 4.666667 |
F7QNwDYG6I | [
"strength: The paper is well written with clear figures that explain the algorithm and process.",
"strength: The algorithm is conceptually interesting, and the comparison of three different techniques to estimate the Q-value function is valuable.",
"strength: The paper is well-structured and written in a clear,... | 4 |
kz5igjl04W | [
"strength: The subject matter itself that this paper deals with is of great scientific importance beyond the toy example.",
"strength: The presentation of this paper is very excellent. It is well thought out for a diverse audience, especially since it provides sufficient background knowledge and motivation for th... | 5.5 |
EyaH1wzmao | [
"strength: The paper claims to have discovered new integer relations between constants that have not been discovered before. This is an exciting and relevant direction for publication in ICLR since machine assisted mathematics has many unrevealed potentials.",
"strength: The methodology is sound and correct, sinc... | 6.333333 |
XT1Cx6cH2a | [
"strength: The authors slightly expand on the insights of the original DAPE paper and provide new results on the original and their improved variant",
"strength: Experiments conducted across two datasets in comparison with multiple popular ‘positional embedding’ methods, including NoPE, RoPE, CoPE, ALiBi, Kerple ... | 5.333333 |
FTpdQBoBd0 | [
"strength: The motivation is clear and the paper is relatively well-written.",
"strength: The idea seems to be intuitive and experiments support this.",
"strength: The authors' analysis of CLIP characterization is very interesting, and focusing on the impact of CLIP characterization on the generated results is ... | 3 |
tIBAOcAvn4 | [
"strength: Originality: The concept of using transfer priors is not entirely new, as it was demonstrated in other works e.g., Dong et al. (2022). However, this paper's novelty lies in applying them to the more practical hard-label setting, where information is limited.",
"strength: Quality: The methodology demons... | 7.5 |
twEvvkQqPS | [
"strength: The originality contribution of this work is excellent by proposing a physics-inspired loss function and model for Hamiltonians prediction problem.",
"strength: Also, the curated dataset is a significant originality contribution by providing a large-scale testbed in the field.",
"strength: The qualit... | 7.5 |
jE8xbmvFin | [
"strength: This paper presents important evidence towards a critical debate (what do LLMs encode).",
"strength: This paper organizes the probing studies in a way that is more comprehensive and rigorous than most probing papers that I have seen. To illustrate that the features are encoded linearly, this paper comp... | 6.75 |
wYvuY60SdD | [
"strength: This work proposes to combine MLP and GNN for balancing the predictions for homophilous and heterophilous nodes.",
"strength: Authors add theoretical analysis that are trying to fold out more insight to show the superiority of the proposed loss.",
"strength: Experiments are conducted on various scale... | 6 |
j7b4mm7Ec9 | [
"strength: Discusses a very noteworthy issue: watermarking of several types of data.",
"strength: The motivation of the paper is grounded in analysis and the experimental results sufficiently validate the claim.",
"strength: They also considered the robustness against diffusion-based attacks.",
"strength: The... | 7.6 |
9qS3HzSDNv | [
"strength: The protein flexibity problem is critical in drug discovery with no good computationally efficient insilico methods. Hence, the proposed method (if works as advertised) could be paradigm shifting.",
"strength: The flow matching components in the paper are generally well explained.",
"strength: The st... | 6.2 |
2Y5Gseybzp | [
"strength: The paper has strong motivation, creating a unified framework for imprecise labels in multi-class classification that can handle different settings of imperfect label information.",
"strength: The general framework is nicely rooted in EM.",
"strength: The proposed framework achieves strong results in... | 6 |
EbCUbPZjM1 | [
"strength: ReGen presents an inverse design approach for generative simulation, which allows for the creation of diverse and complex simulated environments based on agent behavior and textual descriptions.",
"strength: ReGen generates more diverse and complex simulated environments compared to existing simulation... | 5.25 |
6mLzCepPo8 | [
"strength: The overall presentation is fair and easy to follow the key points in the paper.",
"strength: The proposed model is simple yet interesting to some extend. This may bright some insight for researcher in this field.",
"strength: It is a simple method to recognize the node label frequency as important i... | 3.5 |
ak7r4He1qH | [
"strength: The paper presents an approach to evaluating LLMs in clinical environments by introducing AgentClinic, a multimodal agent benchmark. This is a departure from the traditional static question-answering evaluations and provides a more realistic and comprehensive assessment of LLMs' capabilities in medical d... | 7.2 |
2CflgSMLoK | [
"strength: ES shows a reduction in training time without loss in performance, which is promising for computationally expensive tasks.",
"strength: The use of loss evolution for sampling is an interesting approach that addresses the shortcomings of previous static and simple dynamic sampling methods.",
"strength... | 5.75 |
p0DjhjPXl3 | [
"strength: This work focuses on defending against black-box MI attacks which have not been addressed in the existing defense.",
"strength: As a post-processing defense, SSD can be easily integrated with most pre-trained models for defending against black-box attacks.",
"strength: Empirical results demonstrate S... | 6 |
HCCkCjClO0 | [
"strength: The proposed method focuses on the capacity problem of continual learning, which is a core and meaningful side in the long view of CL research. Their approach tries to fix the model capacity by sampling weights rather than storing it, which is creatively motivated from function approximation.",
"streng... | 3 |
A78MiKnGrL | [
"strength: The paper is well-written and organized, making it easy to follow.",
"strength: Leveraging LLM-generated descriptors to enhance VLM performance during inference is a reasonable and meaningful approach.",
"strength: The paper provides notable performance improvements over baseline methods, supporting ... | 4.5 |
h8yg0hT96f | [
"strength: The paper advances the field of BOED by making it more computationally efficient, whilst broadening the scope of problems that can be tackled by BOED by extending it to diffusion-based generative models. The idea of performing joint sampling and optimization using bi-level optimization seems novel to, al... | 7.333333 |
i25WJWnsmq | [
"strength: To the best of my knowledge, using dual-Hawkes process to model disease progression and recovery and using it as RL reward have not been proposed in prior works, so this work has some novelty.",
"strength: Interesting combination of ideas from RL and survival analysis.",
"strength: Experiments conduc... | 3 |
oeLB25A9oO | [
"strength: The main strength of this paper stands in the nice research idea. There have been different works in theoretical machine learning dealing with learned representations in two-layer networks trained with one step of gradient; this manuscript extends findings in the literature introducing novel data distrib... | 3.833333 |
DFSb67ksVr | [
"strength: The problem is significant for many applications.",
"strength: The proposed method is reasonable, and the re-formulation of MCP is insightful.",
"strength: The experiments are solid and sufficient. Analytic experiments such as ablation studies are provided.",
"strength: While prior works have propo... | 6.666667 |
iKgQOAtvsD | [
"strength: The paper is clearly written and motivates the proposed approach well in a lucid manner.",
"strength: The study of making confusing suffixes semantic is very interesting",
"strength: The paper proposes a novel method that \"translates\" these prompts into coherent and human-readable natural language ... | 5.75 |
EVZnnhtMNX | [
"strength: It is desirable to make RLHF use less resources. Though note that this paper does not make DPO use less resources as was claimed in the abstract.",
"strength: The paper addresses the scalability of DPO by offering a more efficient alternative",
"strength: The use of convex optimization reformulation ... | 3 |
mw1PWNSWZP | [
"strength: The new proposed framework that enhances the instruction tuning on LLMs is impressive.",
"strength: The paper is well-organized and well-written with clear motivations, detailed discussion, nice figures, and sufficient comparison experiments, making it easy to follow and understand.",
"strength: This... | 7.333333 |
DgaY5mDdmT | [
"strength: This paper propose an interesting problem that MLLM seems know which areas should be focused but still fails to solve the related problem successfully all the time. By conducting intervention study, the problem is validated.",
"strength: The proposed methods are intuitive and straightforward to the pro... | 7 |
TxIrMD6lAN | [
"strength: The paper is well organised and easy to follow.",
"strength: The authors tried to visualise the results.",
"strength: The paper aims to address a critical issue in incremental learning: the stability-plasticity dilemma. Task-specific adapters, in this context, are a promising approach to balancing st... | 3 |
zSxpnKh1yS | [
"strength: There is a long line of work on unsupervised skill learning based on mutual information maximization between states and skills, most of the work being motivated by intuition and empirical performance. This work complements Eysenbach et al. (2022) by providing a rigorous understanding of the properties of... | 7.5 |
fcJKzwlwcs | [
"strength: The paper is well-written and includes illustrative figures that aid comprehension.",
"strength: It reports that SODP not only surpasses baseline methods in performance but also shows rapid convergence and robustness across diverse tasks and input modalities. Additionally, the included ablation study o... | 4.8 |
kQNlIHlM6m | [
"strength: The presentation of the paper is quite clear and well organized. The example given in the introduction and the first figure help readers understand the problem setup and the challenge.",
"strength: To my knowledge, the problem formulation is novel and the proposed method is original. It introduces grou... | 4.5 |
0fSNU64FV7 | [
"strength: It is interesting to design an ansatz with favorable computational complexity.",
"strength: Novelty: The paper introduces to the best of my knowledge a novel idea to reduce the scaling of the anti-symmetrization process. The idea looks promising, and it is able to recover state-of-the-art results, alth... | 4.25 |
KHkBpvmYVI | [
"strength: Authors identify failure cases in motif scaffolding and draw an informative analogy with the image inverse problem.",
"strength: The authors clearly define the questions they aim to address in their experiments.",
"strength: Indeed, EVA achieves the best performance or is on par with both SOTA sampli... | 6 |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.