Split (2)

train · 2.31k rows

paper_id stringlengths 10 10	items listlengths 4 82	gt float64 1 9.5
xlrpVyMIwz	[ "strength: PE-GQNN combines quantile prediction and distribution recalibration in a single model (not two-stage models), enhancing both predictive accuracy and calibration efficiency.", "strength: By limiting GNN operations to specific features and introducing target values near the output layer, the model effect...	3.75
YZ7NWYBd5z	[ "strength: The Layer-Integrated Channel Attention (LICA) and Scaled Spatial Attention (SSA) mechanisms capture the importance along each channel and at each spatial location, enhancing the detection performance.", "strength: The proposed system outperforms state-of-the-art systems in terms of accuracy and area un...	3
CxXGvKRDnL	[ "strength: Recognizing that diffusion models can be reformulated to use uniform noise, that this allows them to use universal quantization (UQ), and that UQ is an efficient solution to the REC problem for a uniform noise channel.", "strength: Deriving and implementing a \"universally quantized diffusion model\" (...	8
MCNqgUFTHI	[ "strength: This paper introduces a plug-and-play dialogue policy planner with LLMs for proactive learning.", "strength: Empirical results on three datasets show very promising results in both automatic evaluation and human evaluation, and good transfer ability.", "strength: Utilizes a pluggable and fine-tuned d...	6.333333
R7edIYodis	[ "strength: The creation of a large-scale dataset (SIRD-27M) for predicting symbolic integration tasks (Complete Rule Prediction, Rule Prediction, and Integral Prediction) is valuable.", "strength: The integration with SymPy and the efficiency of the proposed guided integral steps (explores 3× fewer branches than ...	3.5
JHoC430Nxi	[ "strength: The strongest aspect of this paper is the diagnostic evaluation in Table 3, which reveals the impact of various design choices.", "strength: Compared to prior art that serves as a baseline implementation, the design choices improve accuracy by ~2 percent (from 96->98 and from 91->93) while reducing lat...	2.5
jKTUlxo5zy	[ "strength: The paper highlights the importance of fine-grained, local regions for image attribution alongside causation of erroneous predictions to image features.", "strength: The novel submodular function proposed in the paper has been demonstrated to outperform several State-of-the-Art (SoTA) approaches.", "...	7.5
enQSCx47Ud	[ "strength: The general concept of using siamese learning + variational inference is interesting and worth exploring. If the claims of robustness and performance hold up under more rigorous evaluation, this could be a very useful approach.", "strength: The paper is well-structured and written clearly (although the...	3
aSoLl0nlzr	[ "strength: The paper studies an interesting bandit problem and the algorithm proposed is, to the best of my knowledge, not studied in the literature.", "strength: Experiments are somewhat extensive and demonstrate the superiorness of CoEBL to the other baselines.", "strength: This work introduces evolutionary a...	4.5
w4C4z80w59	[ "strength: This work addresses a significant issue by reducing the generation of NSFW content.", "strength: Extensive visualizations illustrate the effectiveness of GIE.", "strength: The proposed method doesn't require any fine-tuning concentrated on cross-attention layers -- which is the general methodology in...	6
lo3nlFHOft	[ "strength: new perspectives on an “old” problem by today’s standards (decentralized training), from the point of view of modeling and adaptive optimization", "strength: analytical results are a plus", "strength: the experiments are fairly thorough", "strength: This paper proposes to study an important, impact...	6.666667
WzCEiBILHu	[ "strength: The introduction and background are really a pleasure to read; the SBP is very well introduced a rigorously defined.", "strength: The manuscript present strong theoretical contributions and a novel method to solve the SBP.", "strength: Overall, I found this paper to be a strong contribution. I would ...	7.5
bxfKIYfHyx	[ "strength: The paper is easy to follow in most places, and I like the experiments on the robustness and label efficiency of the reward function.", "strength: The proposed diffusion-based framework demonstrates exceptional performance in decision-making scenarios involving complex dynamics.", "strength: The meth...	7
AcSChDWL6V	[ "strength: The authors provide a novel and insightful comparison between Graph Transformer and MP-GNN (with VN) in both non-uniform and uniform setups. The latter setup, in particular, marks the first time it has been explored from this perspective, revealing that GT and MPGNN-VN do not subsume each other.", "str...	6.5
EUSkm2sVJ6	[ "strength: This paper introduces on a very important problem and gives some solid baselines to tackle it.", "strength: The method, error metrics, experimental settings, baselines, and evaluations are thoughtfully designed (in general).", "strength: the extra mile effort in deriving confidence intervals;", "st...	7.6
qcyn7ESaM8	[ "strength: Understanding class-bias is an important topic critical in several application", "strength: The paper is well-written and easy to follow.", "strength: The paper presents an original approach by showing that PCA-transformed spaces can capture class-level hardness, offering a new perspective on class b...	2.5
ln6QnzBd8o	[ "strength: The paper is well written, with a good structure and presents the mathematics in a clear and coherent way to the reader.", "strength: The proposed minimization of a surrogate function (Sections 3 & 4) is well reinforced with the experiments section, and the motivations in Section 5 are pertinent for su...	4.8
81cta3WQVI	[ "strength: EdgeRunner demonstrates strengths in the mesh tokenization algorithm and the Auto-regressive Auto-encoder framework, which address previous challenges in mesh generation.", "strength: It offers a mesh tokenization algorithm for efficient compression into 1D token sequences, enabling mesh generation wit...	6.5
VyWv7GSh5i	[ "strength: This paper tried to theoretically explain the proposed method.", "strength: The experimental protocol is well described.", "strength: I really enjoyed reading the paper as it offered me several new insights for approaching the IRL problem.", "strength: The approach to tackle the IRL problem by mean...	2.75
y2ch7iQSJu	[ "strength: The selected problem is very relevant and interesting for real life data applications. The paper well states the problem and aspire to provide a solution.", "strength: Authors propose three alternative methods to demonstrate the strength of their approach.", "strength: The use of active learning in s...	2
YSTaRLVP2G	[ "strength: The presentation of the paper is very good.", "strength: All the findings are provided in a digestible and organized way, making the paper an interesting, intriguing and enjoyable read.", "strength: The study is carried out in a systematic way, shedding light into the modus operandi of CNNs.", "str...	4.5
tAcEidZ1Y2	[ "strength: The approach presents an innovative concept and demonstrates commendable performance.", "strength: Notably, the training loss trajectory aligns consistently with that of training on self-supervised MSE, as shown in Figure 6.", "strength: Uncertainty quantification for MRI reconstruction is an open pr...	3.5
RVUWZ9SP1K	[ "strength: By leveraging LLMs to interpret and convert user natural language commands into feasible actions, the ActionFiller framework demonstrates practical potential for complex task automation, aiming to enhance efficiency in task and action sequence planning through experience- and template-based prompting.", ...	3
l9LWx9HMl5	[ "strength: The paper is on a very relevant area and proposes a simple extension to existing approaches for preference optimization.", "strength: The empirical experiments are extensive and instill faith in the reviewer of the proposed algorithms merits.", "strength: The explanation of the unlikelihood objective...	4.8
7nyJBVCTGQ	[ "strength: Overall, the idea seems sufficiently novel and interesting", "strength: The empirical evaluation is extensive and convincing", "strength: Results are state of the art", "strength: The paper is in general well written and I enjoyed reading it. Ideas are explained in detail. For most of the math intu...	7.5
QSuOHV62IQ	[ "strength: The problem that the authors try to attack, i.e. latent causal discovery is crucial in causal discovery. It helps uncover hidden factors influencing observed data, enabling more accurate modeling of causal relationships. Identifying latent variables can improve understanding of complex systems, enhance p...	3.5
jgVqCCg5XX	[ "strength: The paper presents an in-depth evaluation of the proposed MedResEval framework, specifically testing the effects of $N$ (number of parameters) and $D$ (dataset size) — the critical elements of the scaling law. The study defines a formula that effectively models the relationship between performance, param...	4
HgOJlxzB16	[ "strength: The authors analyze an important problem: learning a target function that depends only on two relevant directions with two-layer neural networks. The training algorithm considers joint updates of the first and second layer, this protocol is more closely connected to real-life settings than the layerwise ...	7.5
hv8l922Ad7	[ "strength: The alignment problem is well defined, and even though I am not sure I would agree with the decision, the choice of going for strong disentanglement is well put forward.", "strength: The metrics are fine (up to questions below), and I would expect them to improve on the status quo (which is not satisfa...	3.4
GkJCgUmIqA	[ "strength: The idea of using constraint optimization to solve PINN is interesting.", "strength: The authors gives a detailed description of their method, making it easy to follow.", "strength: Successfully implemented hard constraints within PINNs.", "strength: Mitigated the ill-conditioning issue by alleviat...	3
1NYhrZynvC	[ "strength: This paper uses several examples to demonstrate the benefits of using the proposed step size rules.", "strength: The paper considers a significant and important problem.", "strength: The problem is of current interest -- there are papers appearing about related topics every year.", "strength: The p...	2.5
rFpZnn11gj	[ "strength: Well written introduction", "strength: Good literature survey on LMMs and Pathology--CLIPs", "strength: Brilliant idea on how to develop the revise LLM agent", "strength: Representative patch selection using GPT-4 is an excellent idea", "strength: Interesting approach to select patches through cl...	7.5
skJLOae8ew	[ "strength: Detailed presentation of training details", "strength: Carefully designed data preprocessing procedure for detection and alignment of floor plans", "strength: Reasonable Generation Results for a difficult domain, given architectural designs needs to be coherent and have clear layouts", "strength: T...	3
NN6QHwgRrQ	[ "strength: The proposed constrained formulation to address multiple value alignment problem introduces a novel perspective, and their primal-dual analysis shows an interesting mapping between such formulation and a linear weighted combination formulation. Both theoretical analysis and experiment evaluations are com...	8
jFcNXJGPGh	[ "strength: Dynamic Selector Mechanism: The dynamic selection mechanism is designed to assess the performance of each LoRA component, selecting the best-performing component dynamically. This theoretically enables a more targeted model adaptation, which may contribute to achieving a balance between efficiency and ex...	6
J7ioefqDPw	[ "strength: The paper is well written and well structured", "strength: Pushing for higher standards in adversarial attacks is important for the research community", "strength: The discussion on binary attacks overfitting is interesting (and would worth being investigated more)", "strength: The paper is well-wr...	6.75
sPUrdFGepF	[ "strength: The paper assembles latest advances in dynamic NeRFs and image guided 3D generation to create a novel pipeline for consistent 4D reconstruction where state of the art frame interpolation techniques play a key role in enforcing consistency on single view 3D generation.", "strength: Few architectural inn...	5
tVNZj27pb3	[ "strength: The proposed method is well-motivated.", "strength: The proposed method outperforms the recent baselines", "strength: The authors evaluated the method on both CIL and TIL", "strength: The authors address a crucial issue in continuous learning and pose an intriguing question about the necessity of p...	3.666667
U834XHJuqk	[ "strength: Well written work, with comprehensive introduction, clear problem statement, detailed methodology presentation and thorough experiments. Pseudocode is also given.", "strength: Dimension reduction or representation of sequential data is indeed a crucial problem.", "strength: The method looks novel and...	7.5
WyEdX2R4er	[ "strength: Interesting taskification of data-type identification. I agree with the usefulness for downstream applications such as data curation and data cleaning.", "strength: Introduced TypeIdent dataset spanning 27 data types across 4 categories (geometric, pixel, semantic, and style)", "strength: Extensive e...	8
PDgZ3rvqHn	[ "strength: The proposed method decouples the action dimensions and achieves more flexible and diverse repetition strategies, while maintaining high action persistence.", "strength: The paper provides a thorough comparison with multiple baseline methods, including SAC, TAAC, and UTE. This context clarifies the adv...	6.333333
pppyig2kYe	[ "strength: Unlike standard linear subspace models, LUOS leverages a non-linear latent space to fit complex structures via deep learning methods, which appears effective in capturing non-linear relationships in incomplete data.", "strength: LUOS demonstrates substantial improvements over competing models, achievin...	3
pISLZG7ktL	[ "strength: The focus of this paper, behavior cloning models that generalize to new objects and environments zero shot, has only very recently been shown to work, so analyzing the underlying principles that determine the success of this process is very important.", "strength: The authors approach this problem in a...	8
4LiegvCeQD	[ "strength: The paper addresses a challenging TTA settings where only a single, unlabeled sample is given for adaptation during test time.", "strength: The paper adopts an interesting approach of ensemble learning to dynamically optimize a group of learners (pre-trained models), showing improved TTA performance.",...	2.5
i4ouG6Kc8M	[ "strength: Model selection is a very important and relevant task, especially in the era of the explosion of histopathology foundation models.", "strength: Authors included meaningful tasks beyond histological subtyping, including EGFR prediction and nuclei instance segmentation.", "strength: Originality: The au...	2.5
dxMffCAd4w	[ "strength: The paper presents detailed introduction and explanation of the proposed method, including estimator of the label, optimisation, forward function, initialisation, training process for cases of single node, single layer, and multiple layers. The paper is compressive, well-structured, and easy to follow.",...	2.5
aKRADWBJ1I	[ "strength: Theoretical Rigor: The paper offers a thorough theoretical framework for ACTSAFE, deriving safety guarantees and sample-complexity bounds for safe exploration in continuous state-action spaces—a notable contribution in the safe RL domain.", "strength: Scalability in Practical Settings: The authors exte...	6.75
NQEe7B7bSw	[ "strength: SFT is a timely and impactful area of research as it is an essential phase of the LLM pipeline.", "strength: The proposed GEM method is grounded in theory, effectively tackles overfitting, and improves the output diversity of Llama3-8B.", "strength: The experimental setup is modern.", "strength: Th...	6.25
88JJjsLtqr	[ "strength: The key strengths in this paper are the insights of analyzing different style transfer methods including IP-adpator and Instant Style. With those observations, the proposed method provide a masking solution to demonstrate that removing certain tokens especially the token with high correlations between co...	6.5
NialiwI2V6	[ "strength: This work addresses an important task in medical time-to-event prediction domain and the results are promising.", "strength: The trained model and code is publicly available.", "strength: The experiment results and analysis are comprehensive.", "strength: Paper is well-written and results are inter...	7.5
UhcXE3o1R3	[ "strength: Important Problem - The problem tackled is very important. And it is going to become even more important going forward.", "strength: Method - I overall like the method.", "strength: Method - Method section is written well too.", "strength: The authors introduced a simple but effective approach of u...	4.333333
v0FzmPCd1e	[ "strength: Efficient Memory Management: The Selective Attention mechanism effectively prunes unneeded tokens, significantly reducing memory usage during inference without degrading model performance. This efficiency gain is particularly valuable for scaling transformers in resource-constrained environments.", "st...	6.75
ZO5cn4IfaN	[ "strength: The paper has extensive empirical studies across different learning tasks as well as different network environments.", "strength: The authors also provided a convergence analysis for the proposed method.", "strength: Overall, the proposed method is conceptually simple yet shows promising results.", ...	7
dj8CaE1G7m	[ "strength: Comprehensive Evaluation: Assesses the mechanisms across multiple models (LSTM, VRNN, Transformer) and datasets, demonstrating robust applicability.", "strength: Real-World Relevance: Addresses practical challenges in EHR data, such as irregular time intervals, making findings highly applicable to real...	3
kT0vIJA8CT	[ "strength: This paper novelly uses DDT to learn and represent a reward model, and shows that the method is effective on CartPole, MNIST, and Atari domains.", "strength: Using a tree-based reward model is overall an inspiring approach that is under-explored in the literature.", "strength: Originality: the idea i...	5
mfTM4UdYnC	[ "strength: The paper is clear and straightforward.", "strength: Idea is very cool and could potentially be applied to many domains where reasoning is important. Perhaps even towards mitigating bias and making more fair LLMs.", "strength: Covers many common biases/fallacies.", "strength: simple and straight id...	2.5
Nshk5YpdWE	[ "strength: The paper significantly contributes to the field of neural PDE methods by focusing on enforcing constraints. It marries neural networks with conditioning normalization flows, resulting in LFlows that satisfy the continuity equation by design.", "strength: The method is validated across various 2D and 3...	7.333333
CgpiO0DRrk	[ "strength: In this paper, a novel approach to collaborative edge caching is explored. Rather than concentrating on the sharing of cached content among partners, the paper's primary focus is on the sharing of caching policies among partners. This approach aims to minimize the communication overhead between partners ...	5
IcR1OOFzxm	[ "strength: The motivation of building a generative solver versus a selective one to avoid shortcuts in discriminative reasoning task is very valid and has been highlighted in prior literature [1, 2]. I commend the authors for trying to solve the harder problem in RPM reasoning: mapping the underlying data generativ...	6.5
3ZdGSTxKuy	[ "strength: The paper is well-written and very easy to understand.", "strength: The experimental results of the paper are very good compared to the baseline.", "strength: The first paper to introduce a dataset containing atypical videos in sci-fi and animation category.", "strength: The method tests a strategy...	2
ZRzlhfMqHt	[ "strength: I found the paper to be well written and easy to follow.", "strength: The two introduced attention blocks are interesting and while numerous papers have been published on attention, I believe this work is novel.", "strength: Authors provide extensive empirical evaluation showing superior performance ...	5.25
Ombm8S40zN	[ "strength: The authors introduce the DDPP framework for steering MNDMs by treating the task as sampling from a Bayesian posterior, which is an innovative approach for discrete generative models. They derive three scalable, simulation-free objectives that make fine-tuning computationally efficient, even for large pr...	6.25
7jUQHmz4Tq	[ "strength: This paper proposes a novel dynamic noise injection mechanism that makes sense.", "strength: The performance gains for anomaly localization look good, especially on PRO metric.", "strength: The paper introduces novel contributions to the field of anomaly detection using diffusion models. including dy...	3
I0To0G5J7g	[ "strength: The author’s proposed system doesn’t require a hand-crafted reward signal to do the second fine-tuning stage.", "strength: It is impressive to see how well this works in their experimental results.", "strength: The pipeline also seems straightforward to deploy in practice with the understanding that ...	6.25
n9PDaFNi8t	[ "strength: The work proposes a pretraining-then-fine-tuning framework for GUI grounding and agentic tasks, which is natural and unifies the training and evaluation process of these tasks.", "strength: The work synthesizes a large-scale multi-platform GUI grounding data for pertaining and promises to open-source t...	7.5
FP77VtEuaT	[ "strength: Contrasting previous work, the authors clearly define which notion of \"reasoning\" they are examining and choose a canonical, well-studied classical problem to examine LLM performance on.", "strength: The introduction does a good job of covering most of major, relevant LLM reasoning-related previous w...	5.25
iP8ig954Uz	[ "strength: The approach considers the dependency of the generated parameters across layers, and this is a good observation in my opinion.", "strength: The performance of HART outperforms HyperTuning even though the latter approach requires further pre-training.", "strength: Using contextual conditions to genera...	5.333333
eENHKMTOfW	[ "strength: For the fine-tuning stage of small-scale LLM instructions, by utilizing different types of data and exploring various parameter settings and data organization methods, comprehensive experimental conclusions have been obtained, especially for the setting of large batch size, which is often easily achievab...	6
PhnGhO4VfF	[ "strength: I believe the paper makes valid theoretical contributions which are partially supported experimentally.", "strength: The paper is easy to read and follow and the main takeaway messages easy to understand.", "strength: I believe the studied direction is important to understand the transferability of l...	5.666667
LQdaXixB0g	[ "strength: Originality: This paper makes a meaningful and original contribution by applying sparse autoencoders to uncover latent features in language models specifically related to mental health crises.", "strength: Significance: The author addresses an important and underexplored topic, potentially making a sig...	2.5
F5UgXkPgSn	[ "strength: The problems of subspace clustering and completion of data points on a union of subspaces are of broad interest.", "strength: Numerical experiments on synthetic and real datasets demonstrate that the proposed algorithm can achieve better subspace clustering accuracy when the sampling rate is low.", "...	3
47wXbygsvp	[ "strength: The proposed method can reconstruct an animatable Gaussian avatar without the need of mesh templates.", "strength: The authors propose a strategy to detect \"ambiguous Gaussians\" that may have unreasonable positions or skinning weights. With such a detection strategy, these ambiguous points can be cor...	5
qup9xD8mW4	[ "strength: The introduction of behaviour distillation can enrich the literature and direction of Dataset Distillation. Different from standard DD, behaviour distillation does not require access to the expert datasets. This is quite close to a standard /basic RL setting and could be a good starting point.", "stren...	6.666667
IFOgfaX2Fj	[ "strength: It is a novel application of existing deep learning techniques to the field of orthopaedics.", "strength: Overall empirical scores seem to be quite high, which indicates a good potential of clinical translation.", "strength: GradCAM maps (issue discussed below) seem to provide good qualitative confid...	2.5
n8O0trhost	[ "strength: Instead of finding the deformation for each residue, cryoSPHERE learns a segmentation of the amino acid or nucleotide chains and deforms the segment to fit the heterogenous cryo-EM data. This is indeed a valid assumption in many cases especially for large complexes like the spliceosome (EMPIAR-10180).", ...	6.666667
LvgOm9Rmih	[ "strength: It designs a Visual Grounding task more closely aligned with real-world applications.", "strength: The paper provides a detailed discussion of the process for generating the SG3D dataset, including the specifics of how prompts are designed to generate tasks and the measures taken to minimize potential ...	4.75
7dufGaLYF8	[ "strength: An approach to handling noise in knowledge tracing through two-stream denoising.", "strength: Utilize SVD for explicit and implicit denoising, improving robustness and accuracy.", "strength: Demonstrate performance gains across different standard datasets.", "strength: This paper studies the import...	4
lfRYzd8ady	[ "strength: The authors demonstrate that quantized discrete codebook encodings are more effective representations for continuous control, compared to alternative encodings, such as one-hot and label-based encodings.", "strength: The authors introduce DCWM: Discrete Codebook World Model, a model-based RL method whi...	6.666667
SxOrhLuuVz	[ "strength: The authors motivate their approach by a clever memorization task. They show that vanilla LoRA with low rank (e.g. rank 8) does poorly on a contrived memorization task. They train on 10k unique IDs for 100 epochs where the LLM has to memorize the value associated with a key. This is a nice way to motivat...	4.75
6bKQVm6EOr	[ "strength: The paper is generally well-written and easy to follow. My only comment is about introducing the importance given to the Grassmann manifold in the background and the relatively low importance given to the second term of the loss using it. At first, I was confused, not understanding where the Grassman man...	5.75
fRpAUgKJhT	[ "strength: The motivation is intuitive while reasonable. As far as I know, it is the first work applying the class-aware weighted prompt ensemble.", "strength: The paper is well-written and easy to follow. The authors revisit the existing works, present detailed analyses and explanations, and give clear formulati...	5.75
SThJXvucjQ	[ "strength: The paper is clearly written and mostly easy to follow, with proof roadmap and intuition moderately provided.", "strength: The provided solution nicely connects safe conservative bandits and contextual bandits with general functions class.", "strength: Analysis is sound and rigorous. Numerical experi...	6.666667
tpVQHb4pea	[ "strength: The proposed method is highly scalable, allowing it to create 28 million preferred and rejected response pairs.", "strength: The reward model can improve two different reasoning tasks (math reasoning and logic reasoning).", "strength: This work proposes an interesting idea of preference pre-training....	4.5
sgaMYvGRG5	[ "strength: Interesting uncertainty propagation for geometric completion. The uncertainty exists in almost all computer vision tasks, and it is also very important to the OCC task. In this paper, the authors use the network to predict the uncertainty for each pixel is straightforward. The strategy of computing the u...	5.8
OXIIFZqiiN	[ "strength: Patch analysis is an important topic to study and a field where many ML techniques could potentially be explored.", "strength: The authors also attempt to propose their method from a very theoretical point of view.", "weakness: I highly suspect that the whole paper is generated with some LLMs. See th...	1.5
4KqkizXgXU	[ "strength: This paper continues an interesting line of work in fine-tuning LLMs to be better at red teaming other LLMs. This type of research is complimentary with more recent optimization-based methods like GCG", "strength: The idea of incorporating curiosity into the RL red teaming method from Perez et al. is a...	8
Uj0h13lVrR	[ "strength: The problem this paper seems to try to address is an important one, since deterministic domains is a restrictive assumption and GFlowNets are a promising method.", "strength: The authors address the challenging problem of training a GFlowNet model in a stochastic environment; the approach looks novel."...	1
AxYTFpdlvj	[ "strength: The introduction is very good, as said before. And the idea of using GRDPGs as decoder architecture is interesting, but I think it needs more work.", "strength: The proposed method can be flexibly applied on any existing dot-product-based reconstruction functions.", "strength: Experiments show some i...	2
oBmaLuEJda	[ "strength: This paper demonstrates that large language models (LLMs) can effectively handle semantic understanding tasks with bidirectional self-attention through simple post-training.", "strength: Experimental results on various benchmarks and across different data ratios highlight the effectiveness of BMLM.", ...	3
L8UNn7Llt4	[ "strength: The paper is well-written and easy to follow. The high-level idea is clear with interesting motivation. Theoretical analysis shows the reason harm of backward gradient. The experimental result is also impressive and shows the difference in gradients clearly.", "strength: This paper discusses the relati...	7.5
xizpnYNvQq	[ "strength: Attempts to explain the inner workings of ICL, based on reasonable assumptions and investigative tools.", "strength: The findings align with previous work, encouraging readers to accept the claims presented in the paper.", "strength: Visualized results help readers quickly grasp the core concepts and...	6.5
1aF2D2CPHi	[ "strength: The paper provides a meaningful contribution to open-vocabulary customization for VLMs, especially under data-free constraints.", "strength: It addresses practical issues in adapting CLIP without original data, proposing a unique approach to handle limitations posed by BatchNorm layers.", "strength: ...	8
FowFLhUTgO	[ "strength: Developing high-performance, high-dimensional state-space models is an important topic, and the paper is an interesting attempt in this direction.", "strength: The results in the paper are generally good, demonstrating strong performance on ImageNet classification, COCO detection, and segmentation comp...	5.5
6Mxhg9PtDE	[ "strength: The overall exposition of the issue of depth in safety alignment presented in this paper is quite comprehensive. Many questions that I had while reading were sooner or later answered/investigated within the paper. I particularly also enjoyed how the authors were able to connect fine-tuning attacks into t...	9.5
sYNWqQYJhz	[ "strength: This paper for the first time reveals the vulnerability of safety alignment of LLMs trained via federated instruction tuning.", "strength: This paper proposes the first data poisoning method that aims to compromise the safety of FedIT.", "strength: This paper proposes a post-hoc defense method that e...	5.25
Zes7Wyif8G	[ "strength: First, the paper tackles an important challenge in neuro-symbolic AI, where scalability is still one of the major roadblocks for their general application.", "strength: The achieved speed-ups of the proposed KLay on both CPU and GPU seem very promising.", "strength: Finally, the paper is well-written...	6.5
SjufxrSOYd	[ "strength: This paper introduces Invariant Graphon Networks (IWNs) as an extension of Invariant Graph Networks (IGNs) specifically tailored for graphons, significantly enhancing their expressive power and providing new universal approximation properties for graphon signals. The work advances the theory in graph neu...	8
o4CLLlIaaH	[ "strength: The proposed approach combines geometric priors and neural features to eliminate occlusions in feature-fetching explicitly in the rendering process, which is a novel contribution to the field.", "strength: The authors introduce a nonuniform log sampling strategy and a learnable kernel spatially augment...	6.5
Qqu5mMgIBV	[ "strength: The paper proposes a methodology that fuses the EM algorithm with SEMs to deduce regimes, causal graphs, and functional relationships from observational time-series data.", "strength: The formulation bears resemblance to Rhino and DYNOTEARS, but introduces multiple regimes, providing a new dimension to...	4
UIFAJZ22ZF	[ "strength: The paper used 3D Gaussian Splatting to create large sets of realistic images.", "strength: This method gave them precise control over the test conditions, helping them mitigate confoudning factors.", "strength: Their analysis revealed interesting insights: while DNNs can learn to understand depth fr...	6.666667

Subsets and Splits

No community queries yet

The top public SQL queries from the community will appear here once available.