paper_id string | arxiv_id string | title string | markdown dict | reviews list | scores dict | metadata dict | meta_review dict | decision dict |
|---|---|---|---|---|---|---|---|---|
uy9oR0nYCW | 2410.07436v1 | Toward Robust Real-World Audio Deepfake Detection: Closing the Explainability Gap | {
"content": "## Abstract\n\nAbstract The rapid proliferation of AI-manipulated or generated audio deepfakes poses serious challenges to media integrity and election security. Current AI-driven detection solutions lack explainability and underperform in real-world settings. In this paper, we introduce novel explainab... | [
{
"id": "WS37qBjgYh",
"initial_rating": 1,
"confidence": 4,
"soundness": 2,
"contribution": 1,
"presentation": 1,
"summary": "The paper analyses the existing explainable methods, such as Occlusion and Attention visualization, for deepfake audio detection tasks. The authors show results u... | {
"rating": "1;1;3;5",
"rating_avg": 2.5,
"confidence": "5;4;4;4",
"confidence_avg": 4.25,
"soundness": "1;2;2;3",
"soundness_avg": 2,
"contribution": "1;1;1;2",
"contribution_avg": 1.25,
"presentation": "3;1;2;3",
"presentation_avg": 2.25
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:04.000248"
} | {
"id": "L9VKkhqgoJ",
"metareview": "This paper generally provides an open discussion and analysis of explainable and interpretable algorithms for audio deepfake detection. All the reviewers thought of the low-quality technical contributions. I would like to quote some words from the four reviewers as follows: \n... | {
"decision": "Reject"
} |
uyzkKPvVyS | 2406.16695v2 | Geometry-aware Score Distillation via 3D Consistent Noising and Gradients | {
"content": "## Abstract\n\nAbstract Score distillation sampling (SDS), the methodology in which the score from pretrained 2D diffusion models is distilled into 3D representation, has recently brought significant advancements in text-to-3D generation task. However, this approach is still confronted with critical geo... | [
{
"id": "g1f6Q0hcLE",
"initial_rating": 6,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "This paper focuses on optimization based text-to-3D generation. The authors propose a geometry-aware score distillation method to address the multi-view consisten... | {
"rating": "3;5;5;6",
"rating_avg": 4.75,
"confidence": "4;4;4;3",
"confidence_avg": 3.75,
"soundness": "2;2;2;3",
"soundness_avg": 2.25,
"contribution": "2;2;2;3",
"contribution_avg": 2.25,
"presentation": "1;3;2;3",
"presentation_avg": 2.25
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:04.000948"
} | {
"id": "nCoZxiYEwT",
"metareview": "Summary:\nThe paper aims to improve the geometry consistency of the the text-to-3D generation problem. Its technical contribution lies in a geometry-aware score distillation method, which includes geometry-based gradient warping and gradient dissimilarity loss. \n\n\nStrength:\n... | {
"decision": "Reject"
} |
v0FzmPCd1e | 2410.02703v1 | Selective Attention Improves Transformer | {
"content": "## Abstract\n\nAbstract Unneeded elements in the attention’s context degrade performance.\nWe introduce Selective Attention , a simple parameter-free change to the standard attention mechanism which reduces attention to unneeded elements.\nSelective attention improves language modeling performance in a ... | [
{
"id": "AnVLc5er9k",
"initial_rating": 5,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "The paper introduces a parameter-free modification to the transformer attention mechanism called Selective Attention, which reduces the attention to irrelevant or... | {
"rating": "3;3;5;8",
"rating_avg": 4.75,
"confidence": "4;3;3;4",
"confidence_avg": 3.5,
"soundness": "2;2;3;3",
"soundness_avg": 2.5,
"contribution": "3;3;3;4",
"contribution_avg": 3.25,
"presentation": "3;1;3;3",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:04.001583"
} | {
"id": "Hcy7QhnxjI",
"metareview": "### Summary\nThis paper presents an elegant method for selective attention that prune tokens already absorbed in previous representations. Thus the pruned tokens can be evicted from the kv-cache to improve compute and memory usage.\n\n### Strengths\nThe presented method is intui... | {
"decision": "Accept (Poster)"
} |
v1OQ0kNq0w | 2410.06513v1 | MotionRL: Align Text-to-Motion Generation to Human Preferences with Multi-Reward Reinforcement Learning | {
"content": "## Abstract\n\nAbstract We introduce MotionRL , the first approach to utilize Multi-Reward Reinforcement Learning (RL) for optimizing text-to-motion generation tasks and aligning them with human preferences. Previous works focused on improving numerical performance metrics on the given datasets, often n... | [
{
"id": "6Zeu4whstq",
"initial_rating": 5,
"confidence": 4,
"soundness": 2,
"contribution": 2,
"presentation": 3,
"summary": "The paper proposes incorporating human preference priors into the text-to-motion task to enhance the quality of generated motions. Through the use of reinforcemen... | {
"rating": "5;5;5;6;6",
"rating_avg": 5.4,
"confidence": "4;4;4;3;3",
"confidence_avg": 3.6,
"soundness": "2;3;2;2;3",
"soundness_avg": 2.4,
"contribution": "2;2;2;2;2",
"contribution_avg": 2,
"presentation": "1;3;3;2;3",
"presentation_avg": 2.4
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:04.002182"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
v1f6c7wVBm | 2410.01202v1 | AniSDF: Fused-Granularity Neural Surfaces with Anisotropic Encoding for High-Fidelity 3D Reconstruction | {
"content": "## Abstract\n\nAbstract Neural radiance fields have recently revolutionized novel-view synthesis and achieved high-fidelity renderings.\nHowever, these methods sacrifice the geometry for the rendering quality, limiting their further applications including relighting and deformation.\nHow to synthesize p... | [
{
"id": "jM7dBxEX0Y",
"initial_rating": 5,
"confidence": 4,
"soundness": 2,
"contribution": 3,
"presentation": 3,
"summary": "The paper presents a high-quality 3D surface reconstruction method- AniSDF, which learns fused-granularity neural surfaces with physics-based encoding. The author... | {
"rating": "5;5;6;6;8",
"rating_avg": 6,
"confidence": "4;4;5;4;4",
"confidence_avg": 4.2,
"soundness": "2;2;3;3;3",
"soundness_avg": 2.6,
"contribution": "2;3;2;2;3",
"contribution_avg": 2.4,
"presentation": "2;3;3;3;3",
"presentation_avg": 2.8
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:04.002848"
} | {
"id": "22BFRcbZCd",
"metareview": "The paper proposes a mechanism for high-fidelity 3D reconstruction, showing state-of-the-art results on reconstruction and novel-view synthesis, providing particular gains in view-dependent effects and thin structures, e.g., hair. Reviews are unanimously positive, albeit not all... | {
"decision": "Accept (Poster)"
} |
v1rFkElnIn | 2402.19163v2 | Decoupled Subgraph Federated Learning | {
"content": "## Abstract\n\nAbstract We address the challenge of federated learning on graph-structured data distributed across multiple clients. Specifically, we focus on the prevalent scenario of interconnected subgraphs, where interconnections between different clients play a critical role. We present a novel fra... | [
{
"id": "SUXnwpzzR7",
"initial_rating": 5,
"confidence": 4,
"soundness": 2,
"contribution": 2,
"presentation": 2,
"summary": "The paper works on subgraph FL for node classification, where inter-connections between different clients is important. \n\nIt first computes a global L-hop neigh... | {
"rating": "5;6;6",
"rating_avg": 5.666666666666667,
"confidence": "4;4;4",
"confidence_avg": 4,
"soundness": "2;3;3",
"soundness_avg": 2.6666666666666665,
"contribution": "2;3;3",
"contribution_avg": 2.6666666666666665,
"presentation": "2;4;2",
"presentation_avg": 2.6666666666666665
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:04.004349"
} | {
"id": "xxU7cXLp3Q",
"metareview": "This paper introduces FedStruct, a novel framework for federated learning (FL) on graph-structured data distributed across multiple clients. It focuses on the subgraph FL setting, where interconnections between clients play a significant role. The primary innovation lies in leve... | {
"decision": "Accept (Poster)"
} |
v27yHgKtMv | 2410.15658v1 | Calibration of ordinal regression networks | {
"content": "## Abstract\n\nAbstract Recent studies have shown that deep neural networks are not well-calibrated and produce over-confident predictions.\nThe miscalibration issue primarily stems from the minimization of cross-entropy, which aims to align predicted softmax probabilities with one-hot labels. In ordina... | [
{
"id": "pyP8NcKL9B",
"initial_rating": 5,
"confidence": 5,
"soundness": 2,
"contribution": 2,
"presentation": 2,
"summary": "In this paper, authors propose an approach for calibration of ordinal regression. They propose a loss function that introduces order-aware calibration They use so... | {
"rating": "3;3;5;5",
"rating_avg": 4,
"confidence": "5;4;4;5",
"confidence_avg": 4.5,
"soundness": "1;2;3;2",
"soundness_avg": 2,
"contribution": "2;2;2;2",
"contribution_avg": 2,
"presentation": "1;2;2;2",
"presentation_avg": 1.75
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:04.005176"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
v2zcCDYMok | 2410.05805v1 | PostCast: Generalizable Postprocessing for Precipitation Nowcasting via Unsupervised Blurriness Modeling | {
"content": "## Abstract\n\nAbstract Precipitation nowcasting plays a pivotal role in socioeconomic sectors, especially in severe convective weather warnings.\nAlthough notable progress has been achieved by approaches mining the spatiotemporal correlations with deep learning, these methods still suffer severe blurri... | [
{
"id": "OQaq0JGa4M",
"initial_rating": 3,
"confidence": 3,
"soundness": 2,
"contribution": 2,
"presentation": 2,
"summary": "The paper proposed a universal post-processing method, PostCast, designed for precipitation nowcasting. This method achieves denoising of convolution based predic... | {
"rating": "3;3;6",
"rating_avg": 4,
"confidence": "4;3;3",
"confidence_avg": 3.3333333333333335,
"soundness": "2;2;3",
"soundness_avg": 2.3333333333333335,
"contribution": "2;2;3",
"contribution_avg": 2.3333333333333335,
"presentation": "2;2;3",
"presentation_avg": 2.3333333333333335
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:04.005904"
} | {
"id": "shEYnGkCKZ",
"metareview": "The authors present a method to postprocess blurry precipitation future predictions by using a fine-tuned unconditional diffusion model on precipitation data (to have a reasonable generative prior on precipitation images) and guiding its generation process to the accurate predic... | {
"decision": "Accept (Poster)"
} |
v46TPwU0Uy | 2406.09750v2 | ControlVAR: Exploring Controllable Visual Autoregressive Modeling | {
"content": "## Abstract\n\nAbstract Conditional visual generation has witnessed remarkable progress with the advent of diffusion models (DMs), especially in tasks like control-to-image generation. However, challenges such as expensive computational cost, high inference latency, and difficulties of integration with ... | [
{
"id": "Fy7GUECBME",
"initial_rating": 5,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 1,
"summary": "The paper introduces ControlVAR, a novel framework for controllable visual autoregressive modeling, which integrates pixel-level controls to enhance conditional i... | {
"rating": "3;5;5",
"rating_avg": 4.333333333333333,
"confidence": "5;5;4",
"confidence_avg": 4.666666666666667,
"soundness": "2;2;3",
"soundness_avg": 2.3333333333333335,
"contribution": "1;2;3",
"contribution_avg": 2,
"presentation": "2;2;1",
"presentation_avg": 1.6666666666666667
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:04.006576"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
v4MTnPiYXY | 2411.05193v1 | Q-SFT: Q-Learning for Language Models via Supervised Fine-Tuning | {
"content": "## Abstract\n\nAbstract Value-based reinforcement learning (RL) can in principle learn effective policies for a wide range of multi-turn problems, from games to dialogue to robotic control, including via offline RL from static previously collected datasets. However, despite the widespread use of policy ... | [
{
"id": "lNsnNCpPq1",
"initial_rating": 6,
"confidence": 4,
"soundness": 4,
"contribution": 3,
"presentation": 3,
"summary": "This paper proposed Q-SFT, a method that integrates Q-learning with SFT for LLMs in the multi-turn RL setting. To exploit both the representation and the logits i... | {
"rating": "3;6;6;8",
"rating_avg": 5.75,
"confidence": "4;3;4;4",
"confidence_avg": 3.75,
"soundness": "2;3;4;3",
"soundness_avg": 3,
"contribution": "2;3;3;4",
"contribution_avg": 3,
"presentation": "1;3;3;3",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:04.008010"
} | {
"id": "3YAcYvesaH",
"metareview": "This paper proposes a new way to do RL in multi-turn dialogue setting where the use of RL is less common compared to the vast success of RL in single-turn case. The proposed approach frames Q-learning as a cross-entropy problem with suitably selected weights such that the traine... | {
"decision": "Accept (Poster)"
} |
v5JrYUdMxc | 2405.20669v2 | Hybrid Fourier Score Distillation for Efficient One Image to 3D Object Generation | {
"content": "## Abstract\n\nAbstract Single image-to-3D generation is pivotal for crafting controllable 3D assets. Given its under-constrained nature, we attempt to leverage 3D geometric priors from a novel view diffusion model and 2D appearance priors from an image generation model to guide the optimization process... | [
{
"id": "sbgkDB9geA",
"initial_rating": 5,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 4,
"summary": "The submission proposes a pipeline to generate 3D Gaussian representation from a single image using a score distillation-based method by generating high-frequency... | {
"rating": "3;3;3;5",
"rating_avg": 3.5,
"confidence": "4;5;5;4",
"confidence_avg": 4.5,
"soundness": "2;1;2;3",
"soundness_avg": 2,
"contribution": "2;2;1;3",
"contribution_avg": 2,
"presentation": "3;3;2;4",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:04.008746"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
v8RDgaEtE2 | 2410.05263v1 | Regression Conformal Prediction under Bias | {
"content": "## Abstract\n\nAbstract Uncertainty quantification is crucial to account for the imperfect predictions of machine learning algorithms for high-impact applications. Conformal prediction (CP) is a powerful framework for uncertainty quantification that generates calibrated prediction intervals with valid c... | [
{
"id": "K7RuBcIE17",
"initial_rating": 3,
"confidence": 4,
"soundness": 2,
"contribution": 2,
"presentation": 2,
"summary": "This paper studies the efficiency of conformal prediction when the predicted value has a systematic bias, aiming to understand the effect of symmetric/asymmetric ... | {
"rating": "1;1;3;5",
"rating_avg": 2.5,
"confidence": "3;5;4;3",
"confidence_avg": 3.75,
"soundness": "2;2;2;3",
"soundness_avg": 2.25,
"contribution": "1;1;2;2",
"contribution_avg": 1.5,
"presentation": "2;3;2;3",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:04.009892"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
vDecbmWf6w | 2410.08751v1 | Zero-Shot Offline Imitation Learning via Optimal Transport | {
"content": "## Abstract\n\nAbstract Zero-shot imitation learning algorithms hold the promise of reproducing unseen behavior from as little as a single demonstration at test time.\nExisting practical approaches view the expert demonstration as a sequence of goals, enabling imitation with a high-level goal selector, ... | [
{
"id": "mpMP93iaXx",
"initial_rating": 6,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 4,
"summary": "The authors introduce a new method for zero-shot imitation learning based on optimal transport. \nThe approach involves combining a modified goal-conditioned TD-... | {
"rating": "3;6;6;6",
"rating_avg": 5.25,
"confidence": "3;2;2;3",
"confidence_avg": 2.5,
"soundness": "2;3;3;3",
"soundness_avg": 2.75,
"contribution": "2;4;3;3",
"contribution_avg": 3,
"presentation": "1;4;2;4",
"presentation_avg": 2.75
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:04.011168"
} | {
"id": "fSkUJ2wSmc",
"metareview": "This paper proposes ZILOT (Zero-shot Offline Imitation Learning via Optimal Transport), a novel method for zero-shot imitation learning that aims to address the myopic behavior issue present in existing approaches that decompose expert demonstrations into sequences of individual... | {
"decision": "Reject"
} |
vDp6StrKIq | 2405.15389v2 | Beyond Canonicalization: How Tensorial Messages Improve Equivariant Message Passing | {
"content": "## Abstract\n\nAbstract In many applications of geometric deep learning, the choice of global coordinate frame is arbitrary,\nand predictions should be independent of the reference frame. In other words, the network should be equivariant with respect to rotations and reflections of the input, i.e. the t... | [
{
"id": "xxCLLJFhc5",
"initial_rating": 5,
"confidence": 5,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "This paper focuses on the equivariant message passing and proposes a formalism which together with local canonicalization enables consistent communication of geom... | {
"rating": "5;5;6",
"rating_avg": 5.333333333333333,
"confidence": "3;5;4",
"confidence_avg": 4,
"soundness": "3;3;4",
"soundness_avg": 3.3333333333333335,
"contribution": "2;2;3",
"contribution_avg": 2.3333333333333335,
"presentation": "3;3;3",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:04.011886"
} | {
"id": "wYmFsOytwU",
"metareview": "This paper suggests a way to adapt any message passing architecture to be globally $O(d)$ equivariant by choosing (in an equivariant manner) a local frame at each node, encoding the local $R^d$ feature in this local frame, performing message passing between nodes considering the... | {
"decision": "Accept (Poster)"
} |
vEtDApqkNR | 2405.16440v1 | MambaTS: Improved Selective State Space Models for Long-term Time Series Forecasting | {
"content": "## Abstract\n\nAbstract In recent years, Transformers have become the de-facto architecture for long-term sequence forecasting (LTSF), but faces challenges such as quadratic complexity and permutation invariant bias. A recent model, Mamba, based on selective state space models (SSMs), has emerged as a c... | [
{
"id": "aFbmaLgxxC",
"initial_rating": 5,
"confidence": 2,
"soundness": 2,
"contribution": 2,
"presentation": 2,
"summary": "This paper introduce MambaTS, a new time series forecasting model based on selective state space models. In order to tackle multivariate forecasting, the timeseri... | {
"rating": "3;5;5;6;8",
"rating_avg": 5.4,
"confidence": "4;4;2;3;3",
"confidence_avg": 3.2,
"soundness": "1;2;2;3;3",
"soundness_avg": 2.2,
"contribution": "2;2;2;3;3",
"contribution_avg": 2.4,
"presentation": "3;1;2;2;3",
"presentation_avg": 2.2
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:04.012545"
} | {
"id": "9lTsvomQsa",
"metareview": "The paper introduces MambaTS, a new time series forecasting model based on selective state space models which leverages causal relationships to model global dependencies. Adapting Mamba-style SSMs to the Time Series Space is certainly an interesting direction that is worth inves... | {
"decision": "Reject"
} |
vFVjJsy3PG | 2410.03655v1 | Geometric Representation Condition Improves Equivariant Molecule Generation | {
"content": "## Abstract\n\nAbstract Recent advancements in molecular generative models have demonstrated substantial potential in accelerating scientific discovery, particularly in drug design. However, these models often face challenges in generating high-quality molecules, especially in conditional scenarios wher... | [
{
"id": "aaa8vC8QLZ",
"initial_rating": 5,
"confidence": 3,
"soundness": 2,
"contribution": 2,
"presentation": 3,
"summary": "The paper, titled \"Geometric Representation Condition Improves Equivariant Molecule Generation\" (GeoRCG), presents a novel approach to improving molecular gener... | {
"rating": "5;5;5;5;8",
"rating_avg": 5.6,
"confidence": "4;3;3;3;3",
"confidence_avg": 3.2,
"soundness": "3;4;3;2;3",
"soundness_avg": 3,
"contribution": "3;3;2;2;4",
"contribution_avg": 2.8,
"presentation": "3;4;3;3;3",
"presentation_avg": 3.2
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:04.013341"
} | {
"id": "VvTuO33LPB",
"metareview": "This paper proposes an approach to improve molecule generation by first generating a geometric representation and conditioning the molecule generation on the geometric representation.\n\nThe idea is interesting and shows improvement over the considered EDM architecture. However,... | {
"decision": "Reject"
} |
vFgmobsJiZ | 2406.04344v2 | Verbalized Machine Learning: Revisiting Machine Learning with Language Models | {
"content": "## Abstract\n\nAbstract Motivated by the progress made by large language models (LLMs), we introduce the framework of verbalized machine learning (VML). In contrast to conventional machine learning (ML) models that are typically optimized over a continuous parameter space, VML constrains the parameter s... | [
{
"id": "VZ6jFU4SBj",
"initial_rating": 3,
"confidence": 4,
"soundness": 2,
"contribution": 2,
"presentation": 3,
"summary": "This paper introduces a framework called Verbalized Machine Learning (VML), which constrains the parameter space to human-interpretable natural language. In this ... | {
"rating": "3;3;5;5;6",
"rating_avg": 4.4,
"confidence": "4;4;4;4;5",
"confidence_avg": 4.2,
"soundness": "1;2;2;2;4",
"soundness_avg": 2.2,
"contribution": "2;2;3;2;4",
"contribution_avg": 2.6,
"presentation": "2;3;3;4;4",
"presentation_avg": 3.2
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:04.014520"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
vG123yHVVl | 2312.03419v3 | Synthesizing Physical Backdoor Datasets: An Automated Framework Leveraging Deep Generative Models | {
"content": "## Abstract\n\nAbstract Backdoor attacks, representing an emerging threat to the integrity of deep neural networks, have garnered significant attention due to their ability to compromise deep learning systems clandestinely.\nWhile numerous backdoor attacks occur within the digital realm, their practical... | [
{
"id": "VacPYGzhMm",
"initial_rating": 6,
"confidence": 2,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "The paper presents a framework that can synthesize physical backdoor datasets. The framework consists of three modules: a trigger suggestion module that recommend... | {
"rating": "3;5;5;6",
"rating_avg": 4.75,
"confidence": "5;4;4;2",
"confidence_avg": 3.75,
"soundness": "2;2;3;3",
"soundness_avg": 2.5,
"contribution": "2;2;3;3",
"contribution_avg": 2.5,
"presentation": "1;3;3;3",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:04.015435"
} | {
"id": "H7MW4ziL6Z",
"metareview": "The paper proposed a backdoor framework to generate malicious, physical backdoor dataset based on generative models. The framework involves 3 modules: suggesting the suitable physical triggers, generating the poisoned candidate samples, and finally refining for the most plausibl... | {
"decision": "Reject"
} |
vHO9mU87dc | 2410.21465v1 | ShadowKV: KV Cache in Shadows for High-Throughput Long-Context LLM Inference | {
"content": "## Abstract\n\nAbstract With the widespread deployment of long-context large language models (LLMs), there has been a growing demand for efficient support of high-throughput inference. However, as the key-value (KV) cache expands with the sequence length, the increasing memory footprint and the need to ... | [
{
"id": "BGRFEt6lDF",
"initial_rating": 3,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "This paper presents a high throughput LLM inference system for long sentence length. It proposes a compression method to leverage the low rank property of key cac... | {
"rating": "3;5;5;8",
"rating_avg": 5.25,
"confidence": "4;3;3;3",
"confidence_avg": 3.25,
"soundness": "2;3;2;3",
"soundness_avg": 2.5,
"contribution": "3;3;3;3",
"contribution_avg": 3,
"presentation": "1;3;3;3",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:04.016235"
} | {
"id": "bJXNE2ej2T",
"metareview": "After reviewing the entire discussion and considering input from all reviewers, I recommend the rejection of this paper. While the submission has merits, the unresolved concerns outweigh the strengths. \n\n**Strengths**: \n- The paper presents a novel observation about efficie... | {
"decision": "Reject"
} |
vJ0axKTh7t | 2410.01417v1 | The Labyrinth of Links: Navigating the Associative Maze of Multi-modal LLMs | {
"content": "## Abstract\n\nAbstract Multi-modal Large Language Models (MLLMs) have exhibited impressive capability. However, recently many deficiencies of MLLMs have been found compared to human intelligence, e.g. , hallucination.\nTo drive the MLLMs study, the community dedicated efforts to building larger benchma... | [
{
"id": "komBHzm6RE",
"initial_rating": 6,
"confidence": 5,
"soundness": 3,
"contribution": 3,
"presentation": 2,
"summary": "This work proposes a new benchmark testing the zero-shot association ability of MLLMs. Association origins from object concept learning, where the task is to conn... | {
"rating": "3;5;6;6",
"rating_avg": 5,
"confidence": "4;5;3;5",
"confidence_avg": 4.25,
"soundness": "2;2;3;3",
"soundness_avg": 2.5,
"contribution": "2;3;3;3",
"contribution_avg": 2.75,
"presentation": "1;2;3;2",
"presentation_avg": 2
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:04.016895"
} | {
"id": "JRR0YlAXYs",
"metareview": "This submission introduces a benchmark to test the zero-shot association abilities of multimodal LLMs. The authors argue that the association task, which involves linking observations to prior practice memory, is a fundamental human capability, e.g., connecting images of fresh a... | {
"decision": "Accept (Poster)"
} |
vKG270UOg4 | 2405.17037v1 | BDC-Occ: Binarized Deep Convolution Unit For Binarized Occupancy Network | {
"content": "## Abstract\n\nAbstract Existing 3D occupancy networks demand significant hardware resources, hindering the deployment of edge devices.\nBinarized Neural Networks (BNN) offer substantially reduced computational and memory requirements.\nHowever, their performance decreases notably compared to full-preci... | [
{
"id": "y9b4oPyEph",
"initial_rating": 5,
"confidence": 4,
"soundness": 2,
"contribution": 2,
"presentation": 3,
"summary": "This paper presents a method for binarizing convolution operations in binary occupancy networks. It first analyzes the impact of binarization theoretically, and t... | {
"rating": "3;5;5;5",
"rating_avg": 4.5,
"confidence": "3;2;3;4",
"confidence_avg": 3,
"soundness": "3;3;3;2",
"soundness_avg": 2.75,
"contribution": "3;2;3;2",
"contribution_avg": 2.5,
"presentation": "3;3;2;3",
"presentation_avg": 2.75
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:04.017621"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
vNATZfmY6R | 2407.17773v1 | KiVA: Kid-inspired Visual Analogies for Testing Large Multimodal Models | {
"content": "## Abstract\n\nAbstract This paper investigates visual analogical reasoning in large multimodal models (LMMs) compared to human adults and children. A “visual analogy” is an abstract rule inferred from one image and applied to another.\nWhile benchmarks exist for testing visual reasoning in LMMs, they r... | [
{
"id": "9lh3OXHURG",
"initial_rating": 5,
"confidence": 4,
"soundness": 3,
"contribution": 2,
"presentation": 2,
"summary": "KiVA is a new benchmark for assessing visual analogical reasoning in large multimodal models (LMMs) by comparing their performance to human adults and children. I... | {
"rating": "5;5;8;8",
"rating_avg": 6.5,
"confidence": "3;4;4;4",
"confidence_avg": 3.75,
"soundness": "2;3;3;4",
"soundness_avg": 3,
"contribution": "2;2;4;4",
"contribution_avg": 3,
"presentation": "3;2;4;4",
"presentation_avg": 3.25
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:04.018341"
} | {
"id": "RXa1atVQCr",
"metareview": "Inspired from developmental psychology, the paper introduces a new benchmark to assess visual analogical reasoning in multimodal models and compare it with adults and children. The task involves comparing two images of common household objects (before and after transformation), ... | {
"decision": "Accept (Poster)"
} |
vOFx8HDcvF | 2408.08859v1 | Stochastic Bandits Robust to Adversarial Attacks | {
"content": "## Abstract\n\nAbstract This paper investigates stochastic multi-armed bandit algorithms that are robust to adversarial attacks, where an attacker can first observe the learner’s action and then alter their reward observation.\nWe study two cases of this model, with or without the knowledge of an attack... | [
{
"id": "5CMaTk2zaj",
"initial_rating": 8,
"confidence": 2,
"soundness": 4,
"contribution": 3,
"presentation": 3,
"summary": "The paper develops stochastic multi-armed bandit algorithms that are robust against adversarial attacks. These attacks can alter the reward the learner observes, ... | {
"rating": "6;6;6",
"rating_avg": 6,
"confidence": "4;2;4",
"confidence_avg": 3.3333333333333335,
"soundness": "3;3;3",
"soundness_avg": 3,
"contribution": "2;2;3",
"contribution_avg": 2.3333333333333335,
"presentation": "2;3;3",
"presentation_avg": 2.6666666666666665
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:04.019060"
} | {
"id": "LHfXzB3FAw",
"metareview": "There is a general consensus (although mostly not especially strong) that this work provides sufficiently strong results on robust bandits, particularly distinguishing whether or not the attacker knows the exact action taken. The authors may want to consider mentioning the conc... | {
"decision": "Accept (Poster)"
} |
vPOMTkmSiu | 2402.04177v1 | Scaling Laws for Downstream Task Performance in Machine Translation | {
"content": "## Abstract\n\nAbstract Scaling laws provide important insights that can guide the design of large language models (LLMs). Existing work has primarily focused on studying scaling laws for pretraining (upstream) loss. However, in transfer learning settings, in which LLMs are pretrained on an unsupervised... | [
{
"id": "IuPd6pYxjL",
"initial_rating": 3,
"confidence": 4,
"soundness": 2,
"contribution": 2,
"presentation": 1,
"summary": "The paper proposes scaling laws for pre-training data and downstream data in the context of machine translation. The authors demonstrate that when the pre-trainin... | {
"rating": "3;6;8;8;8",
"rating_avg": 6.6,
"confidence": "4;4;4;3;3",
"confidence_avg": 3.6,
"soundness": "2;3;4;3;3",
"soundness_avg": 3,
"contribution": "2;3;3;3;2",
"contribution_avg": 2.6,
"presentation": "1;2;3;4;3",
"presentation_avg": 2.6
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:04.019783"
} | {
"id": "C2bfpl2Pku",
"metareview": "This paper investigates the scaling laws of the MT downstream performance. More importantly, when pretraining and downstream data are well-aligned, both downstream CE loss and translation quality metrics follow predictable scaling laws; when the distributions are misaligned, the... | {
"decision": "Accept (Poster)"
} |
vQFw9ryKyK | 2410.09874v1 | ImagineNav: Prompting Vision-Language Models as Embodied Navigator through Scene Imagination | {
"content": "## Abstract\n\nAbstract Visual navigation is an essential skill for home-assistance robots, providing the object-searching ability to accomplish long-horizon daily tasks. Many recent approaches use Large Language Models (LLMs) for commonsense inference to improve exploration efficiency. However, the pla... | [
{
"id": "hCa9Jyc2Si",
"initial_rating": 3,
"confidence": 4,
"soundness": 2,
"contribution": 2,
"presentation": 2,
"summary": "This paper proposes \"ImagineNav\", a mapless navigation framework for robots using vision-language models (VLMs) to perform object search in open environments. I... | {
"rating": "3;5;5;6",
"rating_avg": 4.75,
"confidence": "4;4;4;3",
"confidence_avg": 3.75,
"soundness": "2;3;3;3",
"soundness_avg": 2.75,
"contribution": "2;3;2;3",
"contribution_avg": 2.5,
"presentation": "2;3;2;2",
"presentation_avg": 2.25
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:04.020401"
} | {
"id": "38rX2jxHQk",
"metareview": "The paper introduces ImagineNav, a novel mapless visual navigation system for robots using Vision-Language Models (VLMs) to guide object search. The approach replaces traditional mapping and localization with an imagination-driven method that generates and evaluates future views... | {
"decision": "Accept (Poster)"
} |
vQhn4wrQ6j | 2410.01335v1 | Layer Swapping for Zero-Shot Cross-Lingual Transfer in Large Language Models | {
"content": "## Abstract\n\nAbstract Model merging, such as model souping, is the practice of combining different models with the same architecture together without further training. In this work, we present a model merging methodology that addresses the difficulty of fine-tuning Large Language Models (LLMs) for tar... | [
{
"id": "7wEpHwCrsR",
"initial_rating": 8,
"confidence": 4,
"soundness": 4,
"contribution": 3,
"presentation": 4,
"summary": "This paper proposes a new approach to perform zero-shot cross-lingual transfer for solving tasks in a new language. The requirement is to have data for the task i... | {
"rating": "5;6;8",
"rating_avg": 6.333333333333333,
"confidence": "4;4;4",
"confidence_avg": 4,
"soundness": "4;3;4",
"soundness_avg": 3.6666666666666665,
"contribution": "2;3;3",
"contribution_avg": 2.6666666666666665,
"presentation": "3;3;4",
"presentation_avg": 3.3333333333333335
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Spotlight",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:04.021040"
} | {
"id": "9T6DJ9r40c",
"metareview": "The paper introduces a model merging methodology tailored to address the challenge of adapting large language models (LLMs) to perform tasks in non-English languages without task-specific data. The method involves independently fine-tuning \"experts\" on math tasks in English an... | {
"decision": "Accept (Spotlight)"
} |
vQxqcVGrhR | 2410.02067v2 | DisEnvisioner: Disentangled and Enriched Visual Prompt for Customized Image Generation | {
"content": "## Abstract\n\nAbstract In the realm of image generation, creating customized images from visual prompt with additional textual instruction emerges as a promising endeavor. However, existing methods, both tuning-based and tuning-free, struggle with interpreting the subject-essential attributes from the ... | [
{
"id": "pmTFF3FkOt",
"initial_rating": 6,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "This paper introduces DisEnvisioner, a method aimed at enhancing the customization capabilities of image diffusion models. The approach involves training two mode... | {
"rating": "3;5;6;6",
"rating_avg": 5,
"confidence": "4;4;4;3",
"confidence_avg": 3.75,
"soundness": "2;3;3;3",
"soundness_avg": 2.75,
"contribution": "1;3;3;3",
"contribution_avg": 2.5,
"presentation": "2;2;3;3",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:04.021814"
} | {
"id": "W1klUMTJj1",
"metareview": "This paper aims to enhance the customization capabilities of image diffusion models. The authors propose a method DisEnvisioner by training the DIsVisioner and EnVisioner modules to address the limitations of existing methods in interpreting subject-essential attributes. Quantit... | {
"decision": "Accept (Poster)"
} |
vTRWu9zaWo | 2311.08745v5 | Using Stochastic Gradient Descent to Smooth Nonconvex Functions: Analysis of Implicit Graduated Optimization | {
"content": "## Abstract\n\nAbstract The graduated optimization approach is a heuristic method for finding global optimal solutions for nonconvex functions by using a function smoothing operation with stochastic noise. We show that stochastic noise in stochastic gradient descent (SGD) has the effect of smoothing the... | [
{
"id": "03zMzU2q4p",
"initial_rating": 6,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "The paper presents a heuristic approach for solving nonconvex optimization problems by combining a smoothing technique. The authors demonstrate that stochastic gr... | {
"rating": "3;3;5;5;6",
"rating_avg": 4.4,
"confidence": "3;4;3;3;3",
"confidence_avg": 3.2,
"soundness": "3;3;2;3;3",
"soundness_avg": 2.8,
"contribution": "2;1;2;2;3",
"contribution_avg": 2,
"presentation": "3;3;2;4;3",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:04.022636"
} | {
"id": "CXLM7WCqhZ",
"metareview": "The paper examines the smoothing effects of SGD on nonconvex functions, as influenced by the learning rate, batch size, and the variance of the stochastic gradient. While the reviewers found the exploration of the relationship between SGD parameters and smoothing a relevant topi... | {
"decision": "Reject"
} |
vTdwuKUc5Z | 2311.14282v4 | Image Super-Resolution with Text Prompt Diffusion | {
"content": "## Abstract\n\nAbstract Image super-resolution (SR) methods typically model degradation to improve reconstruction accuracy in complex and unknown degradation scenarios. However, extracting degradation information from low-resolution images is challenging, which limits the model performance. To boost ima... | [
{
"id": "F7XCJnWumq",
"initial_rating": 6,
"confidence": 5,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "This paper introduces the text prompts to provide degradation priors for enhancing image SR. Specifically, the authors first develop a text-image generation pipel... | {
"rating": "3;3;5;6",
"rating_avg": 4.25,
"confidence": "5;4;5;5",
"confidence_avg": 4.75,
"soundness": "2;2;2;3",
"soundness_avg": 2.25,
"contribution": "2;1;2;3",
"contribution_avg": 2,
"presentation": "2;2;2;3",
"presentation_avg": 2.25
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:04.023549"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
vVCHWVBsLH | 2410.04907v1 | Decomposition Polyhedra of Piecewise Linear Functions | {
"content": "## Abstract\n\nAbstract In this paper we contribute to the frequently studied question of how to decompose a continuous piecewise linear (CPWL) function into a difference of two convex CPWL functions. Every CPWL function has infinitely many such decompositions, but for applications in optimization and n... | [
{
"id": "SMw3DkYZ2H",
"initial_rating": 5,
"confidence": 3,
"soundness": 3,
"contribution": 2,
"presentation": 2,
"summary": "This paper deals with the following problem: given a continuous piecewise linear (CPWL) function, decompose it as the difference of two convex CPWL functions. Thi... | {
"rating": "5;6;8;8",
"rating_avg": 6.75,
"confidence": "3;3;3;3",
"confidence_avg": 3,
"soundness": "3;3;4;3",
"soundness_avg": 3.25,
"contribution": "2;3;3;3",
"contribution_avg": 2.75,
"presentation": "2;3;4;3",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Spotlight",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:04.024411"
} | {
"id": "F6msTEHbXQ",
"metareview": "The paper studies the problem of decomposing a continuous piecewise-linear (CPWL) function into a difference of two *convex* CPWL functions. The motivation is that CPWL functions can represent all possible ReLU neural networks (and vice-versa), while an explicit representation o... | {
"decision": "Accept (Spotlight)"
} |
vVHc8bGRns | 2410.20868v1 | RecFlow: An Industrial Full Flow Recommendation Dataset | {
"content": "## Abstract\n\nAbstract Industrial recommendation systems (RS) rely on the multi-stage pipeline to balance effectiveness and efficiency when delivering items from a vast corpus to users. Existing RS benchmark datasets primarily focus on the exposure space, where novel RS algorithms are trained and evalu... | [
{
"id": "IB7OUjroCX",
"initial_rating": 6,
"confidence": 4,
"soundness": 3,
"contribution": 4,
"presentation": 2,
"summary": "This paper first proposes a full-flow recommendation dataset collected from the industrial video recommendation scenarios. The overall process includes retrieval,... | {
"rating": "5;6;6;8",
"rating_avg": 6.25,
"confidence": "3;4;4;4",
"confidence_avg": 3.75,
"soundness": "2;3;3;3",
"soundness_avg": 2.75,
"contribution": "3;3;4;4",
"contribution_avg": 3.5,
"presentation": "3;3;2;4",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:04.025439"
} | {
"id": "MgyIpE37VL",
"metareview": "This paper proposes a novel benchmark that includes unexposed items, which is a significant contribution to advancing research on multi-stage recommendation systems. The results from online A/B testing also highlight its practical usefulness.\nThe authors have addressed many of... | {
"decision": "Accept (Poster)"
} |
vVVtTVIR5O | 2410.09365v1 | Debiasing Vison-Language Models with Text-Only Training | {
"content": "## Abstract\n\nAbstract Pre-trained vision-language models (VLMs), such as CLIP, have exhibited remarkable performance across various downstream tasks by aligning text and images in a unified embedding space. However, due to the imbalanced distribution of pre-trained datasets, CLIP suffers from the bias... | [
{
"id": "cQ1VQnSJaA",
"initial_rating": 5,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "The paper addresses debiasing in the context of Vision-Language Models (VLMs). Specifically, the authors argue that existing methods for debiasing VLMs struggle t... | {
"rating": "3;5;5;5",
"rating_avg": 4.5,
"confidence": "3;3;4;4",
"confidence_avg": 3.5,
"soundness": "3;3;3;3",
"soundness_avg": 3,
"contribution": "2;2;2;3",
"contribution_avg": 2.25,
"presentation": "2;2;2;3",
"presentation_avg": 2.25
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:04.026094"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
vVxeFSR4fU | 2406.14479v1 | Tracing Representation Progression: Analyzing and Enhancing Layer-Wise Similarity | {
"content": "## Abstract\n\nAbstract Analyzing the similarity of internal representations within and across different models has been an important technique for understanding the behavior of deep neural networks. Most existing methods for analyzing the similarity between representations of high dimensions, such as t... | [
{
"id": "ACNvhfdbQ8",
"initial_rating": 3,
"confidence": 3,
"soundness": 2,
"contribution": 2,
"presentation": 3,
"summary": "This work studies the feature similarity of neural networks and reveals that: (I) simple-wise cosine similarity can capture representation similarity; (ii) satura... | {
"rating": "3;5;6;6",
"rating_avg": 5,
"confidence": "3;5;2;3",
"confidence_avg": 3.25,
"soundness": "2;3;3;3",
"soundness_avg": 2.75,
"contribution": "2;1;3;3",
"contribution_avg": 2.25,
"presentation": "3;3;3;3",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:04.026938"
} | {
"id": "RzVCzpaiMk",
"metareview": "This paper examines the similarity of internal representations across hidden layers for transformer models. The authors introduce a sample-wise cosine similarity metric and show that enhanced representation similarity leads to increased predicted probability and earlier saturati... | {
"decision": "Accept (Poster)"
} |
vWR3KuiQur | 2411.05007v2 | SVDQuant: Absorbing Outliers by Low-Rank Component for 4-Bit Diffusion Models | {
"content": "## Abstract\n\nAbstract Diffusion models have been proven highly effective at generating high-quality images. However, as these models grow larger, they require significantly more memory and suffer from higher latency, posing substantial challenges for deployment. In this work, we aim to accelerate diff... | [
{
"id": "zC0XuQzUqJ",
"initial_rating": 6,
"confidence": 3,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "This paper presents SVDQuant quantizes weights and activations to 4 bits to accelerate inference of large-scale diffusion models. Through a low-rank decomposition... | {
"rating": "5;6;6;8;8;8",
"rating_avg": 6.833333333333333,
"confidence": "4;3;3;4;4;3",
"confidence_avg": 3.5,
"soundness": "2;3;3;3;4;3",
"soundness_avg": 3,
"contribution": "2;2;2;3;3;4",
"contribution_avg": 2.6666666666666665,
"presentation": "3;2;3;4;4;3",
"presentation_avg": 3.1666666666666665... | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Spotlight",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:04.027680"
} | {
"id": "XbvSUgzxZg",
"metareview": "This paper proposed a novel 4-bit post-quantization paradigm called SVDQuant for diffusion models. It is well written. The experimental parts, a variety of latest diffusion models were evaluated with about 3.5x speedup , demonstrating the universal effectiveness of the proposed ... | {
"decision": "Accept (Spotlight)"
} |
vWRwdmA3wU | 2407.07059v2 | Differentiable Optimization of Similarity Scores Between Models and Brains | {
"content": "## Abstract\n\nAbstract How do we know if two systems – biological or artificial – process information in a similar way? Similarity measures such as linear regression, Centered Kernel Alignment (CKA), Normalized Bures Similarity (NBS), and angular Procrustes distance, are often used to quantify this sim... | [
{
"id": "ZEnb0pOJSq",
"initial_rating": 5,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "This paper studies several popular methods to quantify the similarity between models and neural data by applying them to five neural data from several studies. Th... | {
"rating": "3;5;6;8",
"rating_avg": 5.5,
"confidence": "4;3;3;4",
"confidence_avg": 3.5,
"soundness": "2;3;3;4",
"soundness_avg": 3,
"contribution": "2;3;3;3",
"contribution_avg": 2.75,
"presentation": "2;3;3;3",
"presentation_avg": 2.75
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:04.028470"
} | {
"id": "bIPQl8RBvR",
"metareview": "This paper investigates the properties of a few popular similarity measures and compare them in several neural activity datasets. It addresses the inconsistency across different similarity measures and presents a model-agnostic synthetic dataset optimization to analyze the prope... | {
"decision": "Accept (Poster)"
} |
vcJiPLeC48 | 2410.23467v1 | Gradient-free training of recurrent neural networks | {
"content": "## Abstract\n\nAbstract Recurrent neural networks are a successful neural architecture for many time-dependent problems, including time series analysis, forecasting, and modeling of dynamical systems.\nTraining such networks with\nbackpropagation through time is a notoriously difficult problem because t... | [
{
"id": "ZD7cRWj0U7",
"initial_rating": 5,
"confidence": 5,
"soundness": 4,
"contribution": 2,
"presentation": 2,
"summary": "- The paper proposes a training method for modeling recurrent neural networks without the use of gradient-based methods, such as backpropagation through time (BP... | {
"rating": "5;5;8",
"rating_avg": 6,
"confidence": "4;3;4",
"confidence_avg": 3.6666666666666665,
"soundness": "3;3;4",
"soundness_avg": 3.3333333333333335,
"contribution": "2;3;3",
"contribution_avg": 2.6666666666666665,
"presentation": "3;3;3",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:04.029379"
} | {
"id": "MoPudFgmpc",
"metareview": "In this work, the authors propose a gradient-free approach to training recurrent neural networks in low-dimensions. The proposed approach applies EDMD for output weight training to the sampling RNN framework of Bolager et al. (2023), bypassing backpropagation, and negating the e... | {
"decision": "Reject"
} |
vf5M8YaGPY | 2404.13208v1 | The Instruction Hierarchy: Training LLMs to Prioritize Privileged Instructions | {
"content": "## Abstract\n\nAbstract Today’s LLMs are susceptible to prompt injections, jailbreaks, and other attacks that allow adversaries to overwrite a model’s original instructions with their own malicious prompts.\nIn this work, we argue that one of the primary vulnerabilities underlying these attacks is that ... | [
{
"id": "N1zLsvcKFf",
"initial_rating": 6,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "The authors propose an \"instruction hierarchy\" that enables models to prioritize important instructions over others, ignoring lower-priority or malicious prompt... | {
"rating": "3;5;5;6;6;8;8",
"rating_avg": 5.857142857142857,
"confidence": "4;5;4;5;3;4;4",
"confidence_avg": 4.142857142857143,
"soundness": "1;1;2;3;3;3;3",
"soundness_avg": 2.2857142857142856,
"contribution": "2;3;2;3;3;4;4",
"contribution_avg": 3,
"presentation": "3;4;3;3;3;3;3",
"presentation_... | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:04.030148"
} | {
"id": "qshH1XK9mN",
"metareview": "This paper proposes that some LLM instructions (e.g. those in the system prompt) should override others and implements a data generation method to train models to follow such a hierarchy. The reviewers were conflicted about accepting/rejecting this paper with the most thorough ... | {
"decision": "Reject"
} |
vf5aUZT0Fz | 2410.05021v3 | DEPT: Decoupled Embeddings for Pre-training Language Models | {
"content": "## Abstract\n\nAbstract Language model pre-training benefits from diverse data to enhance performance across domains and languages. However, training on such heterogeneous corpora requires extensive and costly efforts. Since these data sources vary lexically, syntactically, and semantically, they cause ... | [
{
"id": "lztSK9wXgB",
"initial_rating": 5,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 2,
"summary": "The authors introduce the method DEPT, “Decoupled Embeddings for Pre-trained Language models,” an algorithm that trains a transformer model on heterogeneous datas... | {
"rating": "5;6;8",
"rating_avg": 6.333333333333333,
"confidence": "4;4;5",
"confidence_avg": 4.333333333333333,
"soundness": "3;3;4",
"soundness_avg": 3.3333333333333335,
"contribution": "3;4;3",
"contribution_avg": 3.3333333333333335,
"presentation": "2;3;1",
"presentation_avg": 2
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Oral",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:04.031073"
} | {
"id": "VwQoD4Rluw",
"metareview": "This work proposes the DEPT algorithm (Decoupled Embeddings for Pre-Trained language models) to train LMs on diverse data sources more effectively. The method iteratively trains a transformer model on different subsets of data with separate embeddings for each data type (such as... | {
"decision": "Accept (Oral)"
} |
vgV4y086FY | 2409.19800v1 | Differentially Private Bilevel Optimization | {
"content": "## Abstract\n\nAbstract We present differentially private (DP) algorithms for bilevel optimization, a problem class that received significant attention lately in various machine learning applications.\nThese are the first DP algorithms for this task that are able to provide any desired privacy, while al... | [
{
"id": "XlHvIFfst2",
"initial_rating": 6,
"confidence": 2,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "This paper studies bilevel optimization under the central DP model. The authors leverage recent advancements in (non-private) first-order bilevel optimization and... | {
"rating": "5;6;8;8",
"rating_avg": 6.75,
"confidence": "3;2;3;4",
"confidence_avg": 3,
"soundness": "3;3;3;4",
"soundness_avg": 3.25,
"contribution": "2;3;3;3",
"contribution_avg": 2.75,
"presentation": "3;3;4;4",
"presentation_avg": 3.5
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:04.032043"
} | {
"id": "aT9Jms0X50",
"metareview": "Paper studies the Differentially Private optimization of smooth nonconvex-strongly-convex Bilevel problem. Authors provide DP algorithms for solving both the ERM and population level problem in both full-batch and mini-batch settings. Only other work earlier work on DP bilevel o... | {
"decision": "Reject"
} |
vgZDcUetWS | 2406.12816v1 | Neural Approximate Mirror Maps for Constrained Diffusion Models | {
"content": "## Abstract\n\nAbstract Diffusion models excel at creating visually-convincing images, but they often struggle to meet subtle constraints inherent in the training data. Such constraints could be physics-based (e.g., satisfying a PDE), geometric (e.g., respecting symmetry), or semantic (e.g., including a... | [
{
"id": "5ggjwEPeoL",
"initial_rating": 6,
"confidence": 2,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "The paper proposes neural approximate mirror maps for constraint data generation with diffusion models. Compared to typical mirror diffusion models, the mirror ma... | {
"rating": "5;6;6",
"rating_avg": 5.666666666666667,
"confidence": "3;3;2",
"confidence_avg": 2.6666666666666665,
"soundness": "3;3;3",
"soundness_avg": 3,
"contribution": "3;3;3",
"contribution_avg": 3,
"presentation": "2;3;3",
"presentation_avg": 2.6666666666666665
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:04.032996"
} | {
"id": "QA2X3OXSoP",
"metareview": "This paper suggests incorporating constraints into diffusion generative models by mapping samples to an unconstrained space, training a generative diffusion process in the unconstrained space, and then sampling by computing the inverse map of samples in the unconstrained space. ... | {
"decision": "Accept (Poster)"
} |
vh1e2WJfZp | 2410.10105v1 | High-Precision Dichotomous Image Segmentation via Probing Diffusion Capacity | {
"content": "## Abstract\n\nAbstract In the realm of high-resolution (HR), fine-grained image segmentation, the primary challenge is balancing broad contextual awareness with the precision required for detailed object delineation, capturing intricate details and the finest edges of objects.\nDiffusion models, traine... | [
{
"id": "0QbRDNdbDR",
"initial_rating": 5,
"confidence": 3,
"soundness": 3,
"contribution": 2,
"presentation": 2,
"summary": "This paper proposes a model named DiffDIS for high-resolution binary image segmentation, incorporating strategies like single-step denoising, edge-assisted genera... | {
"rating": "5;5;6;6",
"rating_avg": 5.5,
"confidence": "3;3;4;4",
"confidence_avg": 3.5,
"soundness": "3;3;3;4",
"soundness_avg": 3.25,
"contribution": "3;2;3;4",
"contribution_avg": 3,
"presentation": "3;2;3;4",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:04.033640"
} | {
"id": "DDxfs72i6o",
"metareview": "While utilizing diffusion model for image segmentation is not new, this paper proposes a variant of diffusion model for high-precision dichotomous image segmentation. During inference, the method is efficient with single-step denoising from random Gaussian noise. Edge informatio... | {
"decision": "Accept (Poster)"
} |
vhPE3PtTgC | 2410.04456v1 | SWEb: A Large Web Dataset for the Scandinavian Languages | {
"content": "## Abstract\n\nAbstract This paper presents the hitherto largest pretraining dataset for the Scandinavian languages: the Scandinavian WEb (SWEb), comprising over one trillion tokens. The paper details the collection and processing pipeline, and introduces a novel model-based text extractor that signific... | [
{
"id": "Oe3Iqi0KZe",
"initial_rating": 8,
"confidence": 5,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "The paper introduces SWEb, the largest pretraining dataset for Scandinavian languages, containing over one trillion tokens across Swedish, Danish, Norwegian, and ... | {
"rating": "5;5;5;8",
"rating_avg": 5.75,
"confidence": "3;3;4;5",
"confidence_avg": 3.75,
"soundness": "3;4;2;3",
"soundness_avg": 3,
"contribution": "2;3;2;3",
"contribution_avg": 2.5,
"presentation": "3;3;2;3",
"presentation_avg": 2.75
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:04.034288"
} | {
"id": "2ge55nKUeG",
"metareview": "This paper introduces SWEb, a large-scale pretraining dataset consisting of over one trillion tokens for Scandinavian languages, presented with a detailed data collection and processing pipeline including a novel model-based text extractor. Reviewers recognized SWEb's significan... | {
"decision": "Accept (Poster)"
} |
vikwIayXOx | 2409.01062v1 | Random Erasing vs. Model Inversion: A Promising Defense or a False Hope? | {
"content": "## Abstract\n\nAbstract Model Inversion (MI) is a type of privacy violation that focuses on reconstructing private training data through abusive exploitation of machine learning models.\nTo defend against MI attacks,\nstate-of-the-art (SOTA) MI defense methods rely on regularizations that conflict with ... | [
{
"id": "NXrOzdXRsZ",
"initial_rating": 6,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "This paper focuses on defending model inversion (MI) attacks via random erasing. The authors discover that random erasing (RE) has a negative impact on the MI att... | {
"rating": "3;5;5;5;6;6;6",
"rating_avg": 5.142857142857143,
"confidence": "4;4;5;5;3;3;3",
"confidence_avg": 3.857142857142857,
"soundness": "3;3;3;3;3;3;3",
"soundness_avg": 3,
"contribution": "1;2;3;3;2;3;3",
"contribution_avg": 2.4285714285714284,
"presentation": "1;2;2;2;3;3;3",
"presentation_... | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:04.034885"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
vjHySpxDsv | 2410.13726v2 | DAWN: Dynamic Frame Avatar with Non-autoregressive Diffusion Framework for Talking head Video Generation | {
"content": "## Abstract\n\nAbstract Talking head generation intends to produce vivid and realistic talking head videos from a single portrait and speech audio clip. Although significant progress has been made in diffusion-based talking head generation, almost all methods rely on autoregressive strategies, which suf... | [
{
"id": "shqxbrhlRk",
"initial_rating": 6,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "This work presents a non-autoreregrssive diffusion based approach for talking head generation. This allows the generation of videos with non-fixed length. To enha... | {
"rating": "3;5;5;6",
"rating_avg": 4.75,
"confidence": "5;5;5;4",
"confidence_avg": 4.75,
"soundness": "1;3;2;3",
"soundness_avg": 2.25,
"contribution": "2;2;3;3",
"contribution_avg": 2.5,
"presentation": "3;3;3;3",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:04.035570"
} | {
"id": "SWnvhHZ9lw",
"metareview": "This paper received mixed scores: one borderline accept, two borderline reject, and one accept. Most reviewers acknowledged the simple and effective solution proposed in the paper. Specifically, the paper introduces the first non-autoregressive diffusion-based approach for talki... | {
"decision": "Accept (Poster)"
} |
vl8VpW2niQ | 2408.11546v2 | Memorization in In-Context Learning | {
"content": "## Abstract\n\nAbstract In-context learning (ICL) has proven to be an effective strategy for improving the performance of large language models (LLMs) with no additional training. However, the exact mechanism behind these performance improvements remains unclear. This study is the first to show how ICL ... | [
{
"id": "Hi4Eh3La7T",
"initial_rating": 3,
"confidence": 3,
"soundness": 3,
"contribution": 2,
"presentation": 2,
"summary": "The paper investigates the role of memorization in in-context learning (ICL) for LLMs. It shows that ICL significantly surfaces memorized training data, particula... | {
"rating": "3;5;5;6;6",
"rating_avg": 5,
"confidence": "3;3;3;3;4",
"confidence_avg": 3.2,
"soundness": "3;2;3;2;2",
"soundness_avg": 2.4,
"contribution": "2;2;2;2;3",
"contribution_avg": 2.2,
"presentation": "2;3;3;3;4",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:04.036505"
} | {
"id": "evhBdPoo2t",
"metareview": "This paper studies memorization in in-context learning and explores its correlation with downstream performance of LLMs. The key findings suggest that ICL significantly surfaces memorization compared to zero-shot learning, and demonstrations without labels are the effective in ... | {
"decision": "Reject"
} |
vmkpk0ed1F | 2407.00482v1 | Formalizing Spuriousness of Biased Datasets using Partial Information Decomposition | {
"content": "## Abstract\n\nAbstract Spurious patterns refer to a mathematical association between two or more variables in a dataset that are not causally related. However, this notion of spuriousness, which is usually introduced due to sampling biases in the dataset, has classically lacked a formal definition. To ... | [
{
"id": "7kx76zs6U5",
"initial_rating": 3,
"confidence": 4,
"soundness": 2,
"contribution": 1,
"presentation": 2,
"summary": "This paper presents a novel framework to quantify dataset spuriousness, addressing a gap in formalizing how spurious correlations between non-causal features and ... | {
"rating": "3;3;5;6;8",
"rating_avg": 5,
"confidence": "4;4;2;1;2",
"confidence_avg": 2.6,
"soundness": "2;2;3;3;3",
"soundness_avg": 2.6,
"contribution": "1;1;3;3;3",
"contribution_avg": 2.2,
"presentation": "3;2;3;3;3",
"presentation_avg": 2.8
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:04.037347"
} | {
"id": "vBggDRiRvu",
"metareview": "This paper proposes a explanability framework to disentangle spuriousness by using partial information decomposition (PID). The paper's method can isolate four different types of statistical dependencies in a dataset involving the core and spurious features. The paper proposes a... | {
"decision": "Reject"
} |
vo4AHjowKi | 2410.06072v1 | Training-free LLM-generated Text Detection by Mining Token Probability Sequences | {
"content": "## Abstract\n\nAbstract Large language models (LLMs) have demonstrated remarkable capabilities in generating high-quality texts across diverse domains. However, the potential misuse of LLMs has raised significant concerns, underscoring the urgent need for reliable detection of LLM-generated texts. Conve... | [
{
"id": "RRtdzFfizG",
"initial_rating": 5,
"confidence": 5,
"soundness": 3,
"contribution": 3,
"presentation": 4,
"summary": "This paper introduces a novel training-free detector, termed Lastde that synergizes local and global statistics for enhanced detection. They introduce time series... | {
"rating": "5;6;6;6",
"rating_avg": 5.75,
"confidence": "5;3;4;4",
"confidence_avg": 4,
"soundness": "3;3;3;2",
"soundness_avg": 2.75,
"contribution": "3;3;3;2",
"contribution_avg": 2.75,
"presentation": "4;3;3;3",
"presentation_avg": 3.25
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:04.038086"
} | {
"id": "njo8CUo19P",
"metareview": "- Scientific Claims and Findings:\n - The paper introduces a training-free approach for detecting text generated by LLMs. It leverages time series analysis to mine token probability sequences, effectively combining local and global statistics for improved detection accuracy.\... | {
"decision": "Accept (Poster)"
} |
vo9t20wsmd | 2405.19261v2 | Faster Cascades via Speculative Decoding | {
"content": "## Abstract\n\nAbstract Cascades and speculative decoding are two common approaches to improving language models’ inference efficiency.\nBoth approaches\ninvolve\ninterleaving models of different sizes,\nbut via fundamentally distinct mechanisms:\ncascades employ a deferral rule that invokes the larger ... | [
{
"id": "vK9JV4A2i3",
"initial_rating": 3,
"confidence": 4,
"soundness": 1,
"contribution": 1,
"presentation": 3,
"summary": "The paper introduced new speculative decoding variations by combining two techniques: speculative decoding and cascade language models. This is achieved by applyi... | {
"rating": "3;6;8",
"rating_avg": 5.666666666666667,
"confidence": "4;2;3",
"confidence_avg": 3,
"soundness": "1;3;4",
"soundness_avg": 2.6666666666666665,
"contribution": "1;3;3",
"contribution_avg": 2.3333333333333335,
"presentation": "3;4;3",
"presentation_avg": 3.3333333333333335
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Oral",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:04.039025"
} | {
"id": "YBj6R2et11",
"metareview": "This paper proposes speculative cascading, a novel integration of cascading and speculative decoding methods to improve language model inference. The approach intelligently balances speed and quality trade-offs, showing consistent improvements over existing methods under various... | {
"decision": "Accept (Oral)"
} |
vodsIF3o7N | 2410.05656v1 | On the Modeling Capabilities of Large Language Models for Sequential Decision Making | {
"content": "## Abstract\n\nAbstract Large pretrained models are showing increasingly better performance in reasoning and planning tasks across different modalities, opening the possibility to leverage them for complex sequential decision making problems. In this paper, we investigate the capabilities of Large Langu... | [
{
"id": "LVmXIVJld6",
"initial_rating": 5,
"confidence": 3,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "In this paper, the authors study how Large Language Models (LLMs) can produce decision-making policies, either by generating actions or by creating reward models ... | {
"rating": "3;5;5;6",
"rating_avg": 4.75,
"confidence": "4;3;3;4",
"confidence_avg": 3.5,
"soundness": "1;3;2;3",
"soundness_avg": 2.25,
"contribution": "1;2;2;2",
"contribution_avg": 1.75,
"presentation": "3;3;3;3",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:04.039950"
} | {
"id": "toDP7FqYmg",
"metareview": "The paper investigates how large language models can be integrated into reinforcement learning workflows, focusing on two approaches: direct policy generation and indirect reward modeling. Through experiments on diverse environments, the authors demonstrate that LLM-based reward... | {
"decision": "Accept (Poster)"
} |
vpKjmJp6cO | 2407.21260v1 | Bellman Unbiasedness: Toward Provably Efficient Distributional Reinforcement Learning with General Value Function Approximation | {
"content": "## Abstract\n\nAbstract Distributional reinforcement learning improves performance by effectively capturing environmental stochasticity, but a comprehensive theoretical understanding of its effectiveness remains elusive.\nIn this paper, we present a regret analysis for distributional reinforcement learn... | [
{
"id": "2aDVbvfRx4",
"initial_rating": 3,
"confidence": 3,
"soundness": 4,
"contribution": 2,
"presentation": 3,
"summary": "This paper designs a distributional reinforcement learning algorithm with general function approximation, called SF-LSVI, under the assumption that the Eluder dim... | {
"rating": "3;5;6;6;6",
"rating_avg": 5.2,
"confidence": "3;3;3;3;4",
"confidence_avg": 3.2,
"soundness": "4;3;3;4;3",
"soundness_avg": 3.4,
"contribution": "2;2;2;3;3",
"contribution_avg": 2.4,
"presentation": "3;2;2;3;3",
"presentation_avg": 2.6
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:04.040782"
} | {
"id": "B3OghORA3U",
"metareview": "This paper studies distributional reinforcement learning with general value function approximation. The authors design a distributional RL algorithm with general function approximation, called SF-LSVI, and prove that SF-LSVI achieves a near-optimal regret bound under the assump... | {
"decision": "Reject"
} |
vqJZb9SX1T | 2410.09823v1 | Simultaneous Computation and Memory Efficient Zeroth-Order Optimizer for Fine-Tuning Large Language Models | {
"content": "## Abstract\n\nAbstract Fine-tuning is powerful for adapting large language models to downstream tasks, but it often results in huge memory usages.\nA promising approach to mitigate this is using Zeroth-Order (ZO) optimization, which estimates gradients to replace First-Order (FO) gradient calculations,... | [
{
"id": "NB0WuRFpvC",
"initial_rating": 5,
"confidence": 3,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "The authors propose LeZO, a method that integrates the ideas of BCD and ZO-SGD to accelerate training time in comparison to MeZO by Malladi et al. (2023).",
"... | {
"rating": "3;3;5;5",
"rating_avg": 4,
"confidence": "4;4;3;3",
"confidence_avg": 3.5,
"soundness": "2;2;2;3",
"soundness_avg": 2.25,
"contribution": "2;2;3;2",
"contribution_avg": 2.25,
"presentation": "3;2;2;3",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:04.041959"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
vrCT5uCdYp | 2406.05687v2 | FlightBench: Benchmarking Learning-based Methods for Ego-vision-based Quadrotors Navigation | {
"content": "## Abstract\n\nAbstract Ego-vision-based navigation in cluttered environments is crucial for mobile systems, particularly agile quadrotors. While learning-based methods have shown promise recently, head-to-head comparisons with cutting-edge optimization-based approaches are scarce, leaving open the ques... | [
{
"id": "4X5VxWnIkm",
"initial_rating": 3,
"confidence": 3,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "The authors propose FlightBench - a comprehensive benchmark for evaluating ego-vision-based drone navigation methods, comparing learning-based approaches with tra... | {
"rating": "3;5;8",
"rating_avg": 5.333333333333333,
"confidence": "4;3;3",
"confidence_avg": 3.3333333333333335,
"soundness": "3;3;3",
"soundness_avg": 3,
"contribution": "2;2;3",
"contribution_avg": 2.3333333333333335,
"presentation": "3;3;3",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:04.042588"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
vsLohTBH4h | 2401.12526v3 | Refined Generalization Analysis of the Deep Ritz Method and Physics-Informed Neural Networks | {
"content": "## Abstract\n\nAbstract. In this paper, we present refined generalization bounds for the Deep Ritz Method (DRM) and Physics-Informed Neural Networks (PINNs). For the DRM, we focus on two prototype elliptic PDEs: Poisson equation and static Schrödinger equation on the d 𝑑 d italic_d -dimensional unit hy... | [
{
"id": "M91IZgclOi",
"initial_rating": 5,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "Authors utilize the localized analysis to refine generalization bounds for both the Deep Ritz Method (DRM) and Physics-Informed Neural Networks (PINNs). For the D... | {
"rating": "3;5;5;5",
"rating_avg": 4.5,
"confidence": "3;4;4;4",
"confidence_avg": 3.75,
"soundness": "2;3;3;3",
"soundness_avg": 2.75,
"contribution": "1;2;2;3",
"contribution_avg": 2,
"presentation": "1;3;2;3",
"presentation_avg": 2.25
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:04.043868"
} | {
"id": "aZS3wEiuz5",
"metareview": "The manuscript studies the generalization error of machine learning based solvers for partial differential equations. While the work seems solid, overall the reviewers find the novelty marginal compared with existing literature. After reading the manuscript, the metareviewer agr... | {
"decision": "Reject"
} |
vsU2veUpiR | 2410.12949v1 | Mechanistic Unlearning: Robust Knowledge Unlearning and Editing via Mechanistic Localization | {
"content": "## Abstract\n\nAbstract Methods for knowledge editing and unlearning in large language models seek to edit or remove undesirable knowledge or capabilities without compromising general language modeling performance. This work investigates how mechanistic interpretability—which, in part, aims to identify ... | [
{
"id": "NqOcwu8s9y",
"initial_rating": 5,
"confidence": 3,
"soundness": 3,
"contribution": 2,
"presentation": 2,
"summary": "This work explores methods for knowledge editing and unlearning in large language models, focusing on how mechanistic interpretability can enhance the precision a... | {
"rating": "3;3;5;6",
"rating_avg": 4.25,
"confidence": "5;2;3;4",
"confidence_avg": 3.5,
"soundness": "2;3;3;3",
"soundness_avg": 2.75,
"contribution": "2;2;2;4",
"contribution_avg": 2.5,
"presentation": "2;2;2;1",
"presentation_avg": 1.75
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:04.045396"
} | {
"id": "iWiQOxPwQT",
"metareview": "The paper explores the use of mechanistic interpretability to improve knowledge unlearning and editing in LLMs, with a focus on localizing edits. While it demonstrates robustness gains and improved resistance to relearning unwanted information, reviewers highlighted concerns reg... | {
"decision": "Reject"
} |
vue9P1Ypk6 | 2405.12519v1 | MAGE: Model-Level Graph Neural Networks Explanations via Motif-based Graph Generation | {
"content": "## Abstract\n\nAbstract Graph Neural Networks (GNNs) have shown remarkable success in molecular tasks, yet their interpretability remains challenging. Traditional model-level explanation methods like XGNN and GNNInterpreter often fail to identify valid substructures like rings, leading to questionable i... | [
{
"id": "VthXXVZ6JM",
"initial_rating": 5,
"confidence": 4,
"soundness": 2,
"contribution": 3,
"presentation": 3,
"summary": "The paper proposes MAGE, a motif-based explanation method for GNNs in molecular tasks, which identifies significant motifs for each class using an attention-based... | {
"rating": "3;5;5;8",
"rating_avg": 5.25,
"confidence": "3;4;4;4",
"confidence_avg": 3.75,
"soundness": "1;2;2;4",
"soundness_avg": 2.25,
"contribution": "1;2;3;4",
"contribution_avg": 2.5,
"presentation": "2;1;3;3",
"presentation_avg": 2.25
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:04.046533"
} | {
"id": "fViz1TtxDQ",
"metareview": "The paper introduces a motif-based explanation method for Graph Neural Networks (GNNs) applied to molecular tasks. It employs an attention-based learning approach to pinpoint significant motifs for each class. This method provides model-level explanations that highlight crucial ... | {
"decision": "Accept (Poster)"
} |
vx1vJIFvd5 | 2410.11469v1 | O-Edit: Orthogonal Subspace Editing for Language Model Sequential Editing | {
"content": "## Abstract\n\nAbstract Large language models (LLMs) acquire knowledge during pre-training, but over time, this knowledge may become incorrect or outdated, necessitating updates after training. Knowledge editing techniques address this issue without the need for costly re-training. However, most existin... | [
{
"id": "BQkE5SjiQ8",
"initial_rating": 5,
"confidence": 3,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "This paper introduces O-Edit and O-Edit+, two methods for sequential knowledge editing in large language models (LLMs) that address the challenge of catastrophic ... | {
"rating": "5;5;5",
"rating_avg": 5,
"confidence": "4;4;3",
"confidence_avg": 3.6666666666666665,
"soundness": "2;2;3",
"soundness_avg": 2.3333333333333335,
"contribution": "3;2;2",
"contribution_avg": 2.3333333333333335,
"presentation": "2;3;3",
"presentation_avg": 2.6666666666666665
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:04.047835"
} | {
"id": "diDv3DqCV6",
"metareview": "This paper presents O-Edit and O-Edit+, two methods for sequential knowledge editing in large language models (LLMs) that mitigate catastrophic forgetting during multiple edits. The approaches involve projecting update directions into orthogonal subspaces and employing post-pro... | {
"decision": "Reject"
} |
vxBvr5ZpIu | 2410.15336v1 | Diffusion-PINN Sampler | {
"content": "## Abstract\n\nAbstract Recent success of diffusion models has inspired a surge of interest in developing sampling techniques using reverse diffusion processes.\nHowever, accurately estimating the drift term in the reverse stochastic differential equation (SDE) solely from the unnormalized target densit... | [
{
"id": "v6yqppGYvC",
"initial_rating": 3,
"confidence": 5,
"soundness": 2,
"contribution": 2,
"presentation": 2,
"summary": "This paper explores diffusion-based models for unnormalized sampling and introduces the Diffusion-PINN Sampler (DPS). Using physics-informed neural networks (PINN... | {
"rating": "3;3;5;6",
"rating_avg": 4.25,
"confidence": "4;5;4;4",
"confidence_avg": 4.25,
"soundness": "3;2;3;3",
"soundness_avg": 2.75,
"contribution": "2;2;2;3",
"contribution_avg": 2.25,
"presentation": "2;2;3;4",
"presentation_avg": 2.75
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:04.048831"
} | {
"id": "LISBL5biLP",
"metareview": "This paper uses physics-informed neural networks to establish a diffusion-based sampling algorithm. There are two salient concerns from the reviewers about this submission. One is that physics informed neural networks (PINN) have already been used in diffusion-based samplers. It... | {
"decision": "Reject"
} |
vxutwN3xQN | 2407.04842v1 | MJ-Bench: Is Your Multimodal Reward Model Really a Good Judge for Text-to-Image Generation? | {
"content": "## Abstract\n\nAbstract While text-to-image models like DALLE-3 and Stable Diffusion are rapidly proliferating, they often encounter challenges such as hallucination, bias, and the production of unsafe, low-quality output. To effectively address these issues, it is crucial to align these models with des... | [
{
"id": "CrbumVFuJe",
"initial_rating": 6,
"confidence": 4,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "This paper introduces MJ-BENCH, a novel benchmark that includes a comprehensive preference dataset to evaluate multimodal judges in providing feedback for image g... | {
"rating": "5;6;6;6",
"rating_avg": 5.75,
"confidence": "4;4;2;4",
"confidence_avg": 3.5,
"soundness": "2;3;3;3",
"soundness_avg": 2.75,
"contribution": "2;3;1;2",
"contribution_avg": 2,
"presentation": "3;2;3;3",
"presentation_avg": 2.75
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:04.049785"
} | {
"id": "7DpMoM6uha",
"metareview": "Summary of claims and findings: \n\nThe paper introduces a benchmark named MJ-BENCH to evaluate multimodal judges for text-to-image generation models across four critical dimensions: alignment, safety, image quality, and bias. MJ-BENCH includes a comprehensive preference dataset... | {
"decision": "Reject"
} |
vyHFTsOUWu | 2409.14254v1 | Instruction Following without Instruction Tuning | {
"content": "## Abstract\n\nAbstract Instruction tuning commonly means finetuning a language model on instruction-response pairs.\nWe discover two forms of adaptation (tuning) that are deficient compared to instruction tuning, yet still yield instruction following; we call this implicit instruction tuning .\nWe firs... | [
{
"id": "yqhTe0SBHq",
"initial_rating": 3,
"confidence": 4,
"soundness": 2,
"contribution": 2,
"presentation": 2,
"summary": "This paper focuses on exploration of approaches to enable LLM’s instruction following capabilities without instruction tuning. It discusses three different approa... | {
"rating": "3;3;6;10",
"rating_avg": 5.5,
"confidence": "4;4;4;5",
"confidence_avg": 4.25,
"soundness": "2;2;4;4",
"soundness_avg": 3,
"contribution": "2;2;3;4",
"contribution_avg": 2.75,
"presentation": "3;2;4;3",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:04.050564"
} | {
"id": "wNtQC2jFZE",
"metareview": "The paper notes that you can achieve traditional instruction tuning effects by just tuning on responses and then provides a 3 rule method to induce instruction following abilities without actually instruction tuning. The effects of how instruction tuning abilities happen and the... | {
"decision": "Reject"
} |
vyflgpwfJW | 2407.01725v1 | DiscoveryBench: Towards Data-Driven Discovery with Large Language Models | {
"content": "## Abstract\n\nAbstract Can the rapid advances in code generation, function calling, and data analysis using large language models (LLMs) help automate the search and verification of hypotheses purely from a set of provided datasets?\nTo evaluate this question, we present DiscoveryBench , the first comp... | [
{
"id": "uDXqCxK3oO",
"initial_rating": 8,
"confidence": 3,
"soundness": 3,
"contribution": 4,
"presentation": 3,
"summary": "In this paper, the authors consider the question of how capable\nstate-of-the-art LLMs are at automated data-driven discovery. More\nprecisely, the authors presen... | {
"rating": "5;8;8",
"rating_avg": 7,
"confidence": "4;4;3",
"confidence_avg": 3.6666666666666665,
"soundness": "2;4;3",
"soundness_avg": 3,
"contribution": "3;3;4",
"contribution_avg": 3.3333333333333335,
"presentation": "2;3;3",
"presentation_avg": 2.6666666666666665
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:04.051248"
} | {
"id": "le1e7JodBd",
"metareview": "This paper introduces a novel approach to evaluating large language models (LLMs) by leveraging them for data discovery. Two of the reviewers view the paper positively, recognizing it as an important and innovative method for LLM evaluation. However, the third reviewer believes ... | {
"decision": "Accept (Poster)"
} |
vyzPMQ5weJ | 2405.16283v3 | TURNIP: A “Nondeterministic” GPU Runtime with CPU RAM Offload | {
"content": "## Abstract\n\nAbstract An obvious way to alleviate memory difficulties in GPU-based AI computing is via CPU offload , where data are moved between GPU and CPU RAM. While CPU offload is useful, it can greatly slow down a computation due to the relatively slow transfer rate between CPU RAM and GPU RAM. T... | [
{
"id": "VLMh4AcSdY",
"initial_rating": 5,
"confidence": 3,
"soundness": 2,
"contribution": 2,
"presentation": 3,
"summary": "This paper presents TURNIP, a GPU Runtime that incorporates CPU-RAM offload into the execution graph of AI models under memory constraints. Specifically, the auth... | {
"rating": "5;5;5;5;6",
"rating_avg": 5.2,
"confidence": "3;4;3;3;4",
"confidence_avg": 3.4,
"soundness": "3;2;2;2;2",
"soundness_avg": 2.2,
"contribution": "2;2;2;2;3",
"contribution_avg": 2.2,
"presentation": "2;2;2;3;4",
"presentation_avg": 2.6
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:04.051871"
} | {
"id": "MVbb8rGIxk",
"metareview": "Summary\nThe paper proposes a framework to make the inference of large neural networks more memory efficient. The proposed framework, called Turnip, offloads CPU memory by creating MemGraph which is a dependency graph created at compilation time and MemGraph can be used to decid... | {
"decision": "Reject"
} |
vzItLaEoDa | 2410.03618v1 | Open-World Reinforcement Learning over Long Short-Term Imagination | {
"content": "## Abstract\n\nAbstract Training visual reinforcement learning agents in a high-dimensional open world presents significant challenges. While various model-based methods have improved sample efficiency by learning interactive world models, these agents tend to be “short-sighted”, as they are typically t... | [
{
"id": "CQKc8YBsz7",
"initial_rating": 5,
"confidence": 3,
"soundness": 2,
"contribution": 3,
"presentation": 3,
"summary": "The paper introduces a novel hierarchical model-based reinforcement learning (MBRL) framework named Long Short-Term Imagination (LS-Imagine), designed to address ... | {
"rating": "5;6;6;8",
"rating_avg": 6.25,
"confidence": "3;4;4;4",
"confidence_avg": 3.75,
"soundness": "2;3;3;3",
"soundness_avg": 2.75,
"contribution": "3;3;3;4",
"contribution_avg": 3.25,
"presentation": "3;3;3;3",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Oral",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:04.052539"
} | {
"id": "VgZs2AW1Qx",
"metareview": "The paper introduces LS-Image, a model-based RL method that uses hierarchical imagination to solve MineDojo tasks. The key idea is to use a short-term model for step-by-step transitions and a long-term one for multi-step transitions guided by learned affordance maps. These maps ... | {
"decision": "Accept (Oral)"
} |
w0389y0W9D | 2409.14396v1 | Flat-LoRA: Low-Rank Adaption over a Flat Loss Landscape | {
"content": "## Abstract\n\nAbstract Fine-tuning large-scale pre-trained models is prohibitively expensive in terms of computational and memory costs. Low-Rank Adaptation (LoRA), a popular Parameter-Efficient Fine-Tuning (PEFT) method, provides an efficient way to fine-tune models by optimizing only a low-rank matri... | [
{
"id": "4WoI7iNK0I",
"initial_rating": 6,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "This paper introduces Flat-LoRA, a novel extension to the Low-Rank Adaptation (LoRA) framework, designed to optimize model fine-tuning via discovering solutions w... | {
"rating": "5;5;6;6",
"rating_avg": 5.5,
"confidence": "4;4;5;4",
"confidence_avg": 4.25,
"soundness": "3;2;3;3",
"soundness_avg": 2.75,
"contribution": "3;2;3;3",
"contribution_avg": 2.75,
"presentation": "3;3;4;3",
"presentation_avg": 3.25
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:04.053139"
} | {
"id": "AMk5j72Y2F",
"metareview": "Summary: Flat-LoRA adds random weight perturbations into LoRA optimization to locate flat minima in the loss landscape. This improves performance and robustness in vision and language models.\n\nStrengths: novel idea considering full parameter space in LoRA; practical method; cl... | {
"decision": "Reject"
} |
w0b7fCX2nN | 2402.09177v2 | Leveraging the Context through Multi-Round Interactions for Jailbreaking Attacks | {
"content": "## Abstract\n\nAbstract Large Language Models (LLMs) are susceptible to Jailbreaking attacks, which aim to extract harmful information by subtly modifying the attack query. As defense mechanisms evolve, directly obtaining harmful information becomes increasingly challenging for Jailbreaking attacks. In ... | [
{
"id": "g5vhV5Xqx3",
"initial_rating": 3,
"confidence": 3,
"soundness": 2,
"contribution": 2,
"presentation": 2,
"summary": "This paper presents the Contextual Interaction Attack (CIA) against LLMs, which involves prepending a series of prompt, response pairs that gradually lead to a ha... | {
"rating": "3;3;3;6",
"rating_avg": 3.75,
"confidence": "5;5;3;4",
"confidence_avg": 4.25,
"soundness": "2;3;2;3",
"soundness_avg": 2.5,
"contribution": "2;1;2;4",
"contribution_avg": 2.25,
"presentation": "2;3;2;3",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:04.053836"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
w2BELPYbU0 | 2410.07536v2 | I-Max: Maximize the Resolution Potential of Pre-trained Rectified Flow Transformers with Projected Flow | {
"content": "## Abstract\n\nAbstract Rectified Flow Transformers (RFTs) offer superior training and inference efficiency, making them likely the most viable direction for scaling up diffusion models. However, progress in generation resolution has been relatively slow due to data quality and training costs. Tuning-fr... | [
{
"id": "lEu4DMzg50",
"initial_rating": 5,
"confidence": 3,
"soundness": 2,
"contribution": 3,
"presentation": 3,
"summary": "This paper introduces the I-Max framework, designed to address resolution extrapolation in text-to-image models using Rectified Flow Transformers. This paper inco... | {
"rating": "5;5;5;6",
"rating_avg": 5.25,
"confidence": "3;4;3;4",
"confidence_avg": 3.5,
"soundness": "2;3;2;3",
"soundness_avg": 2.5,
"contribution": "3;2;3;3",
"contribution_avg": 2.75,
"presentation": "2;2;3;2",
"presentation_avg": 2.25
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:04.054567"
} | {
"id": "9mME3prlXk",
"metareview": "This paper proposes a tuning-free method to enhance the spatial resolution of text-to-image generation models, grounded on rectified flow transformers. The proposed I-Max framework is thoughtfully designed to improve stability during resolution extrapolation, addressing core que... | {
"decision": "Reject"
} |
w2HYVwXhMh | 2408.03567v1 | Unlocking Exocentric Video-Language Data for Egocentric Video Representation Learning | {
"content": "## Abstract\n\nAbstract We present Embed (Egocentric Models Built with Exocentric Data), a method designed to transform exocentric video-language data for egocentric video representation learning. Large-scale exocentric data covers diverse activities with significant potential for egocentric learning, b... | [
{
"id": "NUdSb7lV3n",
"initial_rating": 5,
"confidence": 4,
"soundness": 2,
"contribution": 2,
"presentation": 3,
"summary": "This paper presents EMBED (Egocentric Models Built with Exocentric Data), a novel approach for adapting exocentric video-language data to improve egocentric video... | {
"rating": "5;5;6;6",
"rating_avg": 5.5,
"confidence": "4;4;5;4",
"confidence_avg": 4.25,
"soundness": "3;2;3;3",
"soundness_avg": 2.75,
"contribution": "2;2;3;3",
"contribution_avg": 2.5,
"presentation": "3;3;3;3",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:04.055182"
} | {
"id": "tgJfKroeTO",
"metareview": "The submission introduces a method to adapt large-scale exocentric videos to train egocentric video understanding models. The main contributions include an algorithm to select and refine close-up views of hand-object interactions, and a language narration rephraser, both of whic... | {
"decision": "Reject"
} |
w2qzdlvPMK | 2411.02592v1 | Decoupled Data Augmentation for Improving Image Classification | {
"content": "## Abstract\n\nAbstract Recent advancements in image mixing and generative data augmentation have shown promise in enhancing image classification. However, these techniques face the challenge of balancing semantic fidelity with diversity. Specifically, image mixing involves interpolating two images to c... | [
{
"id": "1mG4NvU1St",
"initial_rating": 5,
"confidence": 4,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "This paper introduces a method to decouple an image into its class-dependent part (CDP) and class-independent part (CIP), and subsequently processes these two com... | {
"rating": "3;5;5;6",
"rating_avg": 4.75,
"confidence": "4;4;4;4",
"confidence_avg": 4,
"soundness": "2;3;3;3",
"soundness_avg": 2.75,
"contribution": "2;2;2;3",
"contribution_avg": 2.25,
"presentation": "3;2;3;3",
"presentation_avg": 2.75
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:04.055820"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
w2uIJiHTIA | 2404.16676v1 | Multilayer Correlation Clustering | {
"content": "## Abstract\n\nAbstract Correlation Clustering, introduced by Bansal et al. (FOCS ’02), is an elegant optimization model\nthat formulates clustering of objects based on their similarity information.\nIn the model, we are given a set V 𝑉 V italic_V of n 𝑛 n italic_n elements,\nwhere each pair of elemen... | [
{
"id": "oks4O4tEeG",
"initial_rating": 5,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "The paper extends the literature on the fundamental problem of Correlation Clustering. The plot twist here is that there are many correlation clustering instances... | {
"rating": "3;5;5;6",
"rating_avg": 4.75,
"confidence": "3;4;4;3",
"confidence_avg": 3.5,
"soundness": "2;3;3;3",
"soundness_avg": 2.75,
"contribution": "2;3;3;3",
"contribution_avg": 2.75,
"presentation": "3;3;3;3",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:04.056630"
} | {
"id": "ynfMo17iD6",
"metareview": "This paper introduces Multilayer Correlation Clustering, a generalization of the classic Correlation Clustering problem where multiple instances (layers) are defined over the same set of vertices. The goal is to find a single clustering that minimizes disagreements across all la... | {
"decision": "Reject"
} |
w3iM4WLuvy | 2410.08979v2 | Overcoming Slow Decision Frequencies in Continuous Control: Model-Based Sequence Reinforcement Learning for Model-Free Control | {
"content": "## Abstract\n\nAbstract Reinforcement learning (RL) is rapidly reaching and surpassing human-level control capabilities. However, state-of-the-art RL algorithms often require timesteps and reaction times significantly faster than human capabilities, which is impractical in real-world settings and typica... | [
{
"id": "N8CDLWlShl",
"initial_rating": 3,
"confidence": 3,
"soundness": 2,
"contribution": 2,
"presentation": 2,
"summary": "This paper proposes a new model-based RL method that enables a lower frequency for action decision-making, slower than the frequency of sensing and execution. To ... | {
"rating": "3;3;5;6",
"rating_avg": 4.25,
"confidence": "4;3;4;2",
"confidence_avg": 3.25,
"soundness": "1;2;3;3",
"soundness_avg": 2.25,
"contribution": "3;2;2;3",
"contribution_avg": 2.5,
"presentation": "2;2;2;3",
"presentation_avg": 2.25
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:04.057514"
} | {
"id": "pPa8f4AEs1",
"metareview": "This paper addresses the implicit requirement of current RL usage in real-world environments where observations and action frequencies are all assumed to be synchronized with the decision-making algorithm’s frequency, which is difficult to achieve in real time at a high frequenc... | {
"decision": "Accept (Poster)"
} |
w4C4z80w59 | 2408.01014v1 | Growth Inhibitors for Suppressing Inappropriate Image Concepts in Diffusion Models | {
"content": "## Abstract\n\nAbstract. Text-to-image diffusion models have shown the ability to learn a diverse range of concepts. However, it is worth noting that they may also generate undesirable outputs, consequently giving rise to significant security concerns. Specifically, issues such as Not Safe for Work (NSF... | [
{
"id": "sfgwlphjaE",
"initial_rating": 5,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 2,
"summary": "This paper proposes GIE, a method for eliminating NSFW content in diffusion models. The authors leverage the attention map of target concepts, reweighting them to... | {
"rating": "5;5;5;6",
"rating_avg": 5.25,
"confidence": "2;3;4;4",
"confidence_avg": 3.25,
"soundness": "3;2;3;3",
"soundness_avg": 2.75,
"contribution": "3;2;3;4",
"contribution_avg": 3,
"presentation": "3;2;2;3",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:04.058259"
} | {
"id": "uYJHntDzIw",
"metareview": "**Summary of the Paper** \nThis paper introduces GIE, a method for erasing harmful or inappropriate concepts, including implicit NSFW content, from text-to-image diffusion models. Rather than requiring additional fine-tuning, GIE directly modifies the model’s diffusion process... | {
"decision": "Accept (Poster)"
} |
w5ZtXOzMeJ | 2410.03461v1 | Auto-GDA: Automatic Domain Adaptation for Efficient Grounding Verification in Retrieval Augmented Generation | {
"content": "## Abstract\n\nAbstract While retrieval augmented generation (RAG) has been shown to enhance factuality of large language model (LLM) outputs, LLMs still suffer from hallucination, generating incorrect or irrelevant information. One common detection strategy involves prompting the LLM again to assess wh... | [
{
"id": "a1wEgLbpUP",
"initial_rating": 8,
"confidence": 2,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "This paper introduces a novel framework to enhance the performance of NLI models in verifying retrieved evidence within retrieval-augmented generation settings. T... | {
"rating": "3;6;8",
"rating_avg": 5.666666666666667,
"confidence": "4;3;2",
"confidence_avg": 3,
"soundness": "2;3;3",
"soundness_avg": 2.6666666666666665,
"contribution": "2;3;3",
"contribution_avg": 2.6666666666666665,
"presentation": "3;3;3",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:04.059135"
} | {
"id": "3yCRg9xKbW",
"metareview": "This submission aims to address the hallucination problem in RAG systems by enhancing a lightweight NLI model to verify whether the generated outputs are grounded in the retrieved documents. The authors provided a comprehensive rebuttal, effectively addressing most of the review... | {
"decision": "Accept (Poster)"
} |
w6nlcS8Kkn | 2409.12183v2 | To CoT or not to CoT? Chain-of-thought helps mainly on math and symbolic reasoning | {
"content": "## Abstract\n\nAbstract Chain-of-thought (CoT) via prompting is the de facto method for eliciting reasoning capabilities from large language models (LLMs). But for what kinds of tasks is this extra “thinking” really helpful? To analyze this, we conducted a quantitative meta-analysis covering over 100 pa... | [
{
"id": "VNVf0LehMG",
"initial_rating": 6,
"confidence": 4,
"soundness": 3,
"contribution": 2,
"presentation": 4,
"summary": "In this work, the author study the effect of CoT on a wide variety of problems and try to identify categories of problems for which it work v/s ones for which it ... | {
"rating": "6;6;8",
"rating_avg": 6.666666666666667,
"confidence": "4;4;3",
"confidence_avg": 3.6666666666666665,
"soundness": "3;3;4",
"soundness_avg": 3.3333333333333335,
"contribution": "2;2;4",
"contribution_avg": 2.6666666666666665,
"presentation": "2;4;4",
"presentation_avg": 3.33333333333333... | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:04.060406"
} | {
"id": "FFiPcFu7Aw",
"metareview": "This paper presents a comprehensive analysis of chain-of-thought reasoning using large language models. First, the authors present a meta-analysis of existing literature, consisting of 110 papers, and conclude that CoT reasoning mostly benefits tasks in symbolic/logical reasonin... | {
"decision": "Accept (Poster)"
} |
w6rHCuN3YG | 2406.11194v2 | In-Context Editing: Learning Knowledge from Self-Induced Distributions | {
"content": "## Abstract\n\nAbstract In scenarios where language models must incorporate new information efficiently without extensive retraining, traditional fine-tuning methods are prone to overfitting, degraded generalization, and unnatural language generation. To address these limitations, we introduce Consisten... | [
{
"id": "5GQuzOgjFe",
"initial_rating": 8,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "This paper presents ICE, a regularization loss that aims at addressing the limitations of the traditional fine-tuning loss to update knowledge. Experiments on the... | {
"rating": "5;6;6;8",
"rating_avg": 6.25,
"confidence": "4;3;4;3",
"confidence_avg": 3.5,
"soundness": "2;2;3;3",
"soundness_avg": 2.5,
"contribution": "3;3;3;3",
"contribution_avg": 3,
"presentation": "2;4;2;3",
"presentation_avg": 2.75
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:04.061260"
} | {
"id": "8SBxaUBZIl",
"metareview": "This paper introduces an optimization framework for knowledge editing in LLMs. Rather than optimize towards one-hot target distributions, the procedure instead optimizes the model towards a soft distribution induced by conditioning the source model on additional context. This i... | {
"decision": "Accept (Poster)"
} |
w7pMjyjsKN | 2402.01408v2 | Counterfactual Concept Bottleneck Models | {
"content": "## Abstract\n\nAbstract Current deep learning models are not designed to simultaneously address three fundamental questions: predict class labels to solve a given classification task (the \"What?\"), simulate changes in the situation to evaluate how this impacts class predictions (the \"How?\"), and ima... | [
{
"id": "bnIlrMB3Tj",
"initial_rating": 5,
"confidence": 3,
"soundness": 2,
"contribution": 2,
"presentation": 3,
"summary": "The authors extend the concept bottleneck model framework to allow simulation of counterfactual concepts and labels. The method is based on a variational approach... | {
"rating": "5;6;6;8",
"rating_avg": 6.25,
"confidence": "3;4;3;4",
"confidence_avg": 3.5,
"soundness": "2;3;3;3",
"soundness_avg": 2.75,
"contribution": "2;3;3;3",
"contribution_avg": 2.75,
"presentation": "3;4;3;3",
"presentation_avg": 3.25
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:04.062412"
} | {
"id": "qP4BJwPj5f",
"metareview": "The paper extends the concept bottleneck model (CBM) framework to enable counterfactual reasoning, allowing for finer-grained explanations of model predictions using intermediate concepts instead of raw pixel values. The proposed method relies on a variational approach that trea... | {
"decision": "Accept (Poster)"
} |
w8LMtFY97b | 2410.09299v1 | Hierarchical Uncertainty Estimation for Learning-based Registration in Neuroimaging | {
"content": "## Abstract\n\nAbstract Over recent years, deep learning based image registration has achieved impressive accuracy in many domains, including medical imaging and, specifically, human neuroimaging with magnetic resonance imaging (MRI). However, the uncertainty estimation associated with these methods has... | [
{
"id": "YgmihXuCmt",
"initial_rating": 8,
"confidence": 4,
"soundness": 4,
"contribution": 3,
"presentation": 4,
"summary": "Authors present a method to estimate the uncertainty in medical image registration at 3 different stages: (1) on the estimate of the distribution of the deformati... | {
"rating": "3;6;8",
"rating_avg": 5.666666666666667,
"confidence": "4;3;4",
"confidence_avg": 3.6666666666666665,
"soundness": "2;3;4",
"soundness_avg": 3,
"contribution": "1;2;3",
"contribution_avg": 2,
"presentation": "2;3;4",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:04.063230"
} | {
"id": "KLRFYeCSHj",
"metareview": "This work developed a method to estimate the uncertainty in medical image registration at 3 different stages: (1) on the estimate of the distribution of the deformation field at each voxel, (2) on the distribution of the fitted transformation, and (3) on the distribution of poss... | {
"decision": "Accept (Poster)"
} |
wCXAlfvCy6 | 2408.10188v5 | LongVILA: Scaling Long-Context Visual Language Models for Long Videos | {
"content": "## Abstract\n\nAbstract Long-context capability is critical for multi-modal foundation models, especially for long video understanding. We introduce LongVILA, a full-stack solution for long-context visual-language models by co-designing the algorithm and system. For model training, we upgrade existing V... | [
{
"id": "FuVa3dcOds",
"initial_rating": 6,
"confidence": 4,
"soundness": 4,
"contribution": 4,
"presentation": 4,
"summary": "The paper proposes a novel long-context Multi-Modal Sequence Parallelism (MM-SP) system specifically designed to enhance the performance of Vision-Language Models... | {
"rating": "5;6;6",
"rating_avg": 5.666666666666667,
"confidence": "5;4;4",
"confidence_avg": 4.333333333333333,
"soundness": "3;3;4",
"soundness_avg": 3.3333333333333335,
"contribution": "3;3;4",
"contribution_avg": 3.3333333333333335,
"presentation": "3;3;4",
"presentation_avg": 3.333333333333333... | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:04.064056"
} | {
"id": "gqyOLlGgKI",
"metareview": "The paper proposes a novel long-context Multi-Modal Sequence Parallelism system to enhance Vision-Language Models in processing extended contexts, increasing the VILA model's capability from 8 to 1024 frames. Extensive experiments conducted after the rebuttal stage demonstrate t... | {
"decision": "Accept (Poster)"
} |
wD2sfTDy1W | 2406.02780v1 | LADI v2: Multi-label Dataset and Classifiers for Low-Altitude Disaster Imagery | {
"content": "## Abstract\n\nAbstract ML-based computer vision models are promising tools for supporting emergency management operations following natural disasters. Arial photographs taken from small manned and unmanned aircraft can be available soon after a disaster and provide valuable information from multiple pe... | [
{
"id": "PuQb2U51gM",
"initial_rating": 6,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 2,
"summary": "This paper introduces a new multi-label image classification benchmark for natural disasters. During the benchmark construction, the authors extensively involved ... | {
"rating": "3;3;3;6",
"rating_avg": 3.75,
"confidence": "3;5;4;4",
"confidence_avg": 4,
"soundness": "3;2;2;3",
"soundness_avg": 2.5,
"contribution": "2;2;1;3",
"contribution_avg": 2,
"presentation": "3;2;3;2",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:04.064798"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
wDcunIOAOk | 2402.02933v3 | Intrinsic User-Centric Interpretability through Global Mixture of Experts | {
"content": "## Abstract\n\nAbstract Interpretability for neural networks is a trade-off between three key requirements: 1) faithfulness of the explanation ( i.e. , how perfectly it explains the prediction), 2) understandability of the explanation by humans, and 3) model performance. Most existing methods compromise... | [
{
"id": "8qWPDPhDF3",
"initial_rating": 6,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "The paper proposes InterpretCC (ICC), an intrinsic interpretable framework for prediction.\n\nInterpretCC has two modes:\n\n- InterpretCC Feature gating (ICC-FG):... | {
"rating": "5;5;6;8",
"rating_avg": 6,
"confidence": "3;4;3;3",
"confidence_avg": 3.25,
"soundness": "2;3;3;3",
"soundness_avg": 2.75,
"contribution": "2;2;3;3",
"contribution_avg": 2.5,
"presentation": "3;4;3;4",
"presentation_avg": 3.5
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:04.065575"
} | {
"id": "Z0NMlL7V0W",
"metareview": "This paper presents a novel intrinsically interpretable Mixture of Experts (MoE) framework that utilizes two neural networks: one for feature gating and the other for group routing. This approach offers interpretability by highlighting the specific features selected from the ove... | {
"decision": "Accept (Poster)"
} |
wE5xp3zBaQ | 2410.08864v1 | The Good, the Bad and the Ugly: Watermarks, Transferable Attacks and Adversarial Defenses | {
"content": "## Abstract\n\nAbstract We formalize and extend existing definitions of backdoor-based watermarks and adversarial defenses as interactive protocols between two players.\nThe existence of these schemes is inherently tied to the learning tasks for which they are designed.\nOur main result shows that for a... | [
{
"id": "YSqi6hlrav",
"initial_rating": 3,
"confidence": 3,
"soundness": 1,
"contribution": 2,
"presentation": 2,
"summary": "This paper investigates the relationship between watermarks (planted in trained ML models) and adversarial defenses for noiseless classification tasks over a fini... | {
"rating": "3;5;5;6",
"rating_avg": 4.75,
"confidence": "3;4;2;4",
"confidence_avg": 3.25,
"soundness": "1;2;3;3",
"soundness_avg": 2.25,
"contribution": "2;3;2;3",
"contribution_avg": 2.5,
"presentation": "2;2;3;3",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:04.066790"
} | {
"id": "nUA7OSJQYq",
"metareview": "This paper derives a theoretical analysis of the relationship between backdoor-based watermarks and adversarial defenses. The authors show that all discriminative learning tasks can be categorized into one of three classes, which suggests there is an inherent security trade-off ... | {
"decision": "Reject"
} |
wElgE9qBb5 | 2408.06291v1 | Mambular: A Sequential Model for Tabular Deep Learning | {
"content": "## Abstract\n\nAbstract The analysis of tabular data has traditionally been dominated by gradient-boosted decision trees (GBDTs), known for their proficiency with mixed categorical and numerical features. However, recent deep learning innovations are challenging this dominance. We introduce Mambular, an... | [
{
"id": "eUG34zo7S7",
"initial_rating": 3,
"confidence": 5,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "The authors propose Mambular, an adaption of the Mamba architecture for tabular data. The authors compare Mambular against various well-known model families in th... | {
"rating": "3;3;3;6",
"rating_avg": 3.75,
"confidence": "4;3;5;5",
"confidence_avg": 4.25,
"soundness": "1;2;3;3",
"soundness_avg": 2.25,
"contribution": "1;2;3;3",
"contribution_avg": 2.25,
"presentation": "2;2;3;4",
"presentation_avg": 2.75
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:04.068103"
} | {
"id": "FbPf39emtH",
"metareview": "The authors apply the Mamba architecture to tabular data. The proposed technique uses memory more efficiently than transformer-based models and performs about as well as the selected baselines.\n\nHowever, reviewers expressed concerns that evaluation is limited. In addition, no ... | {
"decision": "Reject"
} |
wF9Cz2PknU | 2405.14017v1 | MagicPose4D: Crafting Articulated Models with Appearance and Motion Control | {
"content": "## Abstract\n\nAbstract With the success of 2D and 3D visual generative models, there is growing interest in generating 4D content. Existing methods primarily rely on text prompts to produce 4D content, but they often fall short of accurately defining complex or rare motions. To address this limitation,... | [
{
"id": "Hy4R4qgfPi",
"initial_rating": 5,
"confidence": 4,
"soundness": 4,
"contribution": 3,
"presentation": 4,
"summary": "This paper proposes a novel framework for 4D generation, which can accpet video as input and generate consistent motion for 3D mesh. To be more specific, the dual... | {
"rating": "3;5;5;6",
"rating_avg": 4.75,
"confidence": "5;3;4;4",
"confidence_avg": 4,
"soundness": "2;3;4;3",
"soundness_avg": 3,
"contribution": "2;3;3;2",
"contribution_avg": 2.5,
"presentation": "3;2;4;3",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:04.068774"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
wFD16gwpze | 2410.09005v1 | Analyzing Neural Scaling Laws in Two-Layer Networks with Power-Law Data Spectra | {
"content": "## Abstract\n\nAbstract Neural scaling laws describe how the performance of deep neural networks scales with key factors such as training data size, model complexity, and training time, often following power-law behaviors over multiple orders of magnitude. Despite their empirical observation, the theore... | [
{
"id": "243eGtVGn7",
"initial_rating": 6,
"confidence": 4,
"soundness": 4,
"contribution": 3,
"presentation": 3,
"summary": "This paper studies the learning dynamics of committee machines trained on random data with power law covariance structure. They utilize a hierarchy of order param... | {
"rating": "6;6;6",
"rating_avg": 6,
"confidence": "3;3;4",
"confidence_avg": 3.3333333333333335,
"soundness": "3;4;4",
"soundness_avg": 3.6666666666666665,
"contribution": "2;2;3",
"contribution_avg": 2.3333333333333335,
"presentation": "3;4;3",
"presentation_avg": 3.3333333333333335
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Spotlight",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:04.069634"
} | {
"id": "Zt3EaLKq8d",
"metareview": "This work investigates the presence of neural scaling laws in a two-layer student-teacher setup under the assumption that the data has a prescribed number of distinct eigenvalues from a power-law distribution, and in the limit of large data and dimension. Reviewers found the pap... | {
"decision": "Accept (Spotlight)"
} |
wFs2E5wCw6 | 2410.11201v1 | Tree of Attributes Prompt Learning for Vision-Language Models | {
"content": "## Abstract\n\nAbstract Prompt learning has proven effective in adapting vision language models for downstream tasks. However, existing methods usually append learnable prompt tokens solely with the category names to obtain textual features, which fails to fully leverage the rich context indicated in th... | [
{
"id": "iCgvZjmX1A",
"initial_rating": 5,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "- The paper proposes a new prompt learning approach for vision language models. \n- Instead of a fixed prompt or LLM-generated unstructured prompt, the paper prop... | {
"rating": "5;5;5;6;8",
"rating_avg": 5.8,
"confidence": "5;1;3;4;5",
"confidence_avg": 3.6,
"soundness": "2;2;3;3;4",
"soundness_avg": 2.8,
"contribution": "2;2;3;3;4",
"contribution_avg": 2.8,
"presentation": "3;3;3;3;3",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:04.070547"
} | {
"id": "LboqUI6SOC",
"metareview": "This paper proposes the Tree of Attributes Prompt learning (TAP) by first instructing LLMs to generate a tree of attributes with a “concept - attribute - description” structure for each category, followed other learning the hierarchy with vision and text prompt tokens. Initially... | {
"decision": "Accept (Poster)"
} |
wGVOxplEbf | 2409.06633v1 | SaRA: High-Efficient Diffusion Model Fine-tuning with Progressive Sparse Low-Rank Adaptation | {
"content": "## Abstract\n\nAbstract In recent years, the development of diffusion models has led to significant progress in image, video, and 3D generation tasks, with pre-trained models like the Stable Diffusion series playing a crucial role.\nHowever, a key challenge remains in downstream task applications: how t... | [
{
"id": "z3j0zHGYd3",
"initial_rating": 5,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "The paper presents a novel approach for fine-tuning pre-trained diffusion models called SaRA for visual content generation. The method builds on a key insight: pa... | {
"rating": "5;5;5;5;8",
"rating_avg": 5.6,
"confidence": "2;3;4;3;5",
"confidence_avg": 3.4,
"soundness": "2;2;2;3;4",
"soundness_avg": 2.6,
"contribution": "2;2;3;3;4",
"contribution_avg": 2.8,
"presentation": "3;3;3;3;4",
"presentation_avg": 3.2
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:04.071588"
} | {
"id": "eXoqr5Kuul",
"metareview": "This paper works on the fine-tuning of diffusion models for new tasks. The basic idea is to identify the importance of parameters in the pre-trained diffusion model as the parameters with smallest absolute values. Then the proposed method fine-tunes these ineffective parameters ... | {
"decision": "Accept (Poster)"
} |
wGqf7YMF8R | 2409.17433v1 | HDFlow: Enhancing LLM Complex Problem-Solving with Hybrid Thinking and Dynamic Workflows | {
"content": "## Abstract\n\nAbstract Despite recent advancements in large language models (LLMs), their performance on complex reasoning problems requiring multi-step thinking and combining various skills is still limited. To address this, we propose a novel framework HDFlow for complex reasoning with LLMs that comb... | [
{
"id": "7nkQTF0cqg",
"initial_rating": 3,
"confidence": 4,
"soundness": 2,
"contribution": 2,
"presentation": 3,
"summary": "This paper presents a fast-slow thinking mechanism where the fast thinking is direct CoT and slow thinking is a dynamic workflow method. It also utilizes a datase... | {
"rating": "3;5;5;6",
"rating_avg": 4.75,
"confidence": "4;5;4;4",
"confidence_avg": 4.25,
"soundness": "2;3;3;4",
"soundness_avg": 3,
"contribution": "2;3;2;4",
"contribution_avg": 2.75,
"presentation": "3;3;3;4",
"presentation_avg": 3.25
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:04.072491"
} | {
"id": "qnd5oWd3OZ",
"metareview": "This paper proposes a new method for LLM reasoning. Specificallly, the authors propose a new approach for slow, deliberate reasoning called Dynamic Workflow, which automatically decomposes complex problems into more manageable sub-tasks and dynamically designs a workflow to asse... | {
"decision": "Reject"
} |
wH8XXUOUZU | 2410.10733v2 | Deep Compression Autoencoder for Efficient High-Resolution Diffusion Models | {
"content": "## Abstract\n\nAbstract We present Deep Compression Autoencoder (DC-AE), a new family of autoencoders for accelerating high-resolution diffusion models. Existing autoencoders have demonstrated impressive results at a moderate spatial compression ratio (e.g., 8 × \\times × ), but fail to maintain satisfa... | [
{
"id": "oweKmsgNXI",
"initial_rating": 6,
"confidence": 4,
"soundness": 2,
"contribution": 2,
"presentation": 2,
"summary": "This paper presents a family of autoencoders that achieve comparable or better reconstructions compared to the SD autoencoders while having a much stronger spatia... | {
"rating": "5;6;6;6;8",
"rating_avg": 6.2,
"confidence": "4;4;4;4;2",
"confidence_avg": 3.6,
"soundness": "4;3;3;2;3",
"soundness_avg": 3,
"contribution": "3;3;3;2;3",
"contribution_avg": 2.8,
"presentation": "4;2;3;2;4",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:04.073044"
} | {
"id": "yzNpHdvA21",
"metareview": "This paper explores image auto-encoders with a high spatial compression rate in the context of image generation. The proposed model results in efficiency gains while maintaining downstream performance. The manuscript was reviewed by four knowledgeable reviewers who acknowledged ... | {
"decision": "Accept (Poster)"
} |
wHLMsM1SrP | 2411.05000v1 | Needle Threading: Can LLMs Follow Threads Through Near-Million-Scale Haystacks? | {
"content": "## Abstract\n\nAbstract As the context limits of Large Language Models (LLMs) increase, the range of possible applications and downstream functions broadens. In many real-world tasks, decisions depend on details scattered across collections of often disparate documents containing mostly irrelevant infor... | [
{
"id": "Jjkr0BcQ9I",
"initial_rating": 6,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "In this paper, the authors introduce simple single-needle retrieval, multiple-needle and conditional-needle retrieval and challenging needle threading and multith... | {
"rating": "5;6;8",
"rating_avg": 6.333333333333333,
"confidence": "2;4;3",
"confidence_avg": 3,
"soundness": "3;3;3",
"soundness_avg": 3,
"contribution": "3;3;3",
"contribution_avg": 3,
"presentation": "2;3;4",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:04.073568"
} | {
"id": "3V2Qnxm5HV",
"metareview": "This paper focuses on experimental validation and discussion of the context understanding capabilities of existing LLMs. The authors designed a series of retrieval experiments, including single needles, and multiple needles with random and clustered distributions. The experiment... | {
"decision": "Accept (Poster)"
} |
wI5uHZLeCZ | 2407.15549v2 | Latent Adversarial Training Improves Robustness to Persistent Harmful Behaviors in LLMs | {
"content": "## Abstract\n\nAbstract Large language models (LLMs) can often be made to behave in undesirable ways that they are explicitly fine-tuned not to.\nFor example, the LLM red-teaming literature has produced a wide variety of ‘jailbreaking’ techniques to elicit harmful text from models that were fine-tuned t... | [
{
"id": "JaR2UYkcZV",
"initial_rating": 5,
"confidence": 4,
"soundness": 2,
"contribution": 2,
"presentation": 2,
"summary": "This paper introduces targeted latent adversarial training (LAT) as a technique to improve robustness to persistent harmful behaviors in large language models (LL... | {
"rating": "3;3;3;5",
"rating_avg": 3.5,
"confidence": "5;4;3;4",
"confidence_avg": 4,
"soundness": "3;3;2;2",
"soundness_avg": 2.5,
"contribution": "3;2;2;2",
"contribution_avg": 2.25,
"presentation": "2;1;2;2",
"presentation_avg": 1.75
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:04.074288"
} | {
"id": "pB7r1t2ibX",
"metareview": "This paper introduces targeted latent adversarial training (LAT) as a technique to improve robustness to persistent harmful behaviors in large language models (LLMs). The authors demonstrate LAT's effectiveness in three key applications: (1) improving resistance to jailbreaking ... | {
"decision": "Reject"
} |
wJPMe9UKow | 2406.00410v1 | Posterior Label Smoothing for Node Classification | {
"content": "## Abstract\n\nAbstract Soft labels can improve the generalization of a neural network classifier in many domains, such as image classification. Despite its success, the current literature has overlooked the efficiency of label smoothing in node classification with graph-structured data.\nIn this work, ... | [
{
"id": "cOKvr11VuU",
"initial_rating": 3,
"confidence": 4,
"soundness": 3,
"contribution": 2,
"presentation": 2,
"summary": "The paper introduces a method called Posterior Label Smoothing (PosteL) designed to improve node classification tasks in graph-structured data. By integrating bot... | {
"rating": "3;5;5;8",
"rating_avg": 5.25,
"confidence": "4;4;1;3",
"confidence_avg": 3,
"soundness": "3;2;3;3",
"soundness_avg": 2.75,
"contribution": "2;2;3;2",
"contribution_avg": 2.25,
"presentation": "2;4;3;3",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:04.075301"
} | {
"id": "zpoCGwl9St",
"metareview": "This paper proposes a label smoothing method for transductive node classification, which incorporates local context through neighborhood label distribution. However, the reviewers provided negative feedback, raising several concerns, including: (1) the assumption that neighborho... | {
"decision": "Reject"
} |
wJv4AIt4sK | 2405.20935v1 | Effective Interplay between Sparsity and Quantization: From Theory to Practice | {
"content": "## Abstract\n\nAbstract The increasing size of deep neural networks necessitates effective model compression to improve computational efficiency and reduce their memory footprint.\nSparsity and quantization are two prominent compression methods that have individually demonstrated significant reduction i... | [
{
"id": "lDdVaHFtOv",
"initial_rating": 8,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 4,
"summary": "The paper explores the relationship between two widely used compression techniques: sparsity and quantization. Specifically, it demonstrates that these techniques... | {
"rating": "5;6;8;8",
"rating_avg": 6.75,
"confidence": "4;4;4;4",
"confidence_avg": 4,
"soundness": "2;2;4;3",
"soundness_avg": 2.75,
"contribution": "3;3;3;3",
"contribution_avg": 3,
"presentation": "3;3;4;4",
"presentation_avg": 3.5
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Spotlight",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:04.076039"
} | {
"id": "Zp6RtEOAft",
"metareview": "In this paper the authors consider the error introduced by various model compression schemes for deep networks, such as sparsifying or quantizing the numerical precision of the weights. While both approaches are commonly used, here the authors consider the combination of the tw... | {
"decision": "Accept (Spotlight)"
} |
wLmJIs1uqG | 2404.03331v1 | LancBiO: Dynamic Lanczos-aided Bilevel Optimization via Krylov Subspace | {
"content": "## Abstract\n\nAbstract Bilevel optimization, with broad applications in machine learning, has an intricate hierarchical structure. Gradient-based methods have emerged as a common approach to large-scale bilevel problems. However, the computation of the hyper-gradient, which involves a Hessian inverse v... | [
{
"id": "XaqZbm43BX",
"initial_rating": 5,
"confidence": 4,
"soundness": 3,
"contribution": 2,
"presentation": 2,
"summary": "This paper investigates a gradient based method for solving bilevel optimization problems with a strongly convex lower-level objective. To enhance the efficiency ... | {
"rating": "5;5;5;6",
"rating_avg": 5.25,
"confidence": "3;3;4;3",
"confidence_avg": 3.25,
"soundness": "2;3;3;3",
"soundness_avg": 2.75,
"contribution": "2;2;2;3",
"contribution_avg": 2.25,
"presentation": "3;3;2;3",
"presentation_avg": 2.75
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:04.077271"
} | {
"id": "LxBjLaYbCG",
"metareview": "This paper investigates bilevel optimization problems with a strongly convex lower-level objective, introducing novel algorithms, SubBiO and LancBiO, that leverage Krylov subspace methods and the Lanczos process to efficiently approximate hypergradients without directly computin... | {
"decision": "Accept (Poster)"
} |
wLzhEQq2hR | 2410.00193v1 | Do Vision-Language Models Really Understand Visual Language? | {
"content": "## Abstract\n\nAbstract Visual language is a system of communication that conveys information through symbols, shapes, and spatial arrangements. Diagrams are a typical example of a visual language depicting complex concepts and their relationships in the form of an image. The symbolic nature of diagrams... | [
{
"id": "tRfenM6Gff",
"initial_rating": 5,
"confidence": 4,
"soundness": 3,
"contribution": 2,
"presentation": 2,
"summary": "The paper demonstrates an in-depth evaluation if Vision Language models (VLM) can understand visual digram. The authors show the results both on their synthetical... | {
"rating": "5;5;6;8",
"rating_avg": 6,
"confidence": "2;4;2;3",
"confidence_avg": 2.75,
"soundness": "3;3;3;4",
"soundness_avg": 3.25,
"contribution": "2;2;3;3",
"contribution_avg": 2.5,
"presentation": "3;2;3;4",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:04.078737"
} | {
"id": "3gLfrjHUqg",
"metareview": "The paper presents an evaluation of whether VLMs can understand visual diagrams. While VLMs recognize and reason about entities, they struggle with reasoning about relations in synthetic diagrams. Surprisingly, their performance on understanding relations improves on more comple... | {
"decision": "Reject"
} |
wMRFTQwp1d | 2407.06491v1 | VideoEval: Comprehensive Benchmark Suite for Low-Cost Evaluation of Video Foundation Model | {
"content": "## Abstract\n\nAbstract With the growth of high-quality data and advancement in visual pre-training paradigms, Video Foundation Models (VFMs) have made significant progress recently, demonstrating their remarkable performance on traditional video understanding benchmarks. However, the existing benchmark... | [
{
"id": "ZeBMFTBIk2",
"initial_rating": 3,
"confidence": 4,
"soundness": 3,
"contribution": 2,
"presentation": 2,
"summary": "The paper proposes an ensemble of video datasets for foundation model benchmarking, and provides different evaluation protocols. The proposed benchmarks consist o... | {
"rating": "3;3;5;5",
"rating_avg": 4,
"confidence": "5;4;4;4",
"confidence_avg": 4.25,
"soundness": "2;3;3;2",
"soundness_avg": 2.5,
"contribution": "1;2;2;3",
"contribution_avg": 2,
"presentation": "3;2;3;4",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:04.079754"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
wN3KaUXA5X | 2405.20519v1 | Diffusion On Syntax Trees For Program Synthesis | {
"content": "## Abstract\n\nAbstract Large language models generate code one token at a time. Their autoregressive generation process lacks the feedback of observing the program’s output. Training LLMs to suggest edits directly can be challenging due to the scarcity of rich edit data. To address these problems, we p... | [
{
"id": "47gavgaNRT",
"initial_rating": 8,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 4,
"summary": "The authors propose a program synthesis method based on \"tree diffusion\". They randomly corrupt programs (with some constraints) and learn to invert the corrupt... | {
"rating": "6;6;6;8;8",
"rating_avg": 6.8,
"confidence": "3;3;3;5;4",
"confidence_avg": 3.6,
"soundness": "3;3;4;3;3",
"soundness_avg": 3.2,
"contribution": "2;3;3;3;3",
"contribution_avg": 2.8,
"presentation": "3;2;3;4;4",
"presentation_avg": 3.2
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Spotlight",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:04.080434"
} | {
"id": "HdPvUQo9tr",
"metareview": "This paper introduces a novel diffusion-based approach to program synthesis, leveraging syntax trees to iteratively refine programs while preserving syntactic validity. The authors position their method as an alternative to traditional autoregressive models, which are limited by... | {
"decision": "Accept (Spotlight)"
} |
wNg0LibmQt | 2410.03489v2 | Gradient-based Jailbreak Images for Multimodal Fusion Models | {
"content": "## Abstract\n\nAbstract Augmenting language models with image inputs may enable more effective jailbreak attacks through continuous optimization, unlike text inputs that require discrete optimization. However, new multimodal fusion models tokenize all input modalities using non-differentiable functions,... | [
{
"id": "dz56N96fCI",
"initial_rating": 8,
"confidence": 4,
"soundness": 3,
"contribution": 4,
"presentation": 4,
"summary": "* The paper develops a white box gradient-based image jailbreak method for multimodal fusion models. Prior work on gradient-based image jailbreaks has focused on ... | {
"rating": "3;3;3;6;8",
"rating_avg": 4.6,
"confidence": "4;4;3;3;4",
"confidence_avg": 3.6,
"soundness": "1;3;2;3;3",
"soundness_avg": 2.4,
"contribution": "1;2;1;3;4",
"contribution_avg": 2.2,
"presentation": "1;3;3;3;4",
"presentation_avg": 2.8
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:04.081025"
} | {
"id": "ScXQ7RJNvC",
"metareview": "This paper presents a method for white-box gradient-based image jailbreak attacks on multimodal fusion models using a \"tokenizer shortcut\" for continuous optimization. The proposed method achieves higher attack success rates than text-based jailbreaks and is more computational... | {
"decision": "Reject"
} |
wP0nDEAlap | 2312.00591v1 | Less is More: Learning Reference Knowledge Using No-Reference Image Quality Assessment | {
"content": "## Abstract\n\nAbstract Image Quality Assessment (IQA) with reference images have achieved great success by imitating the human vision system, in which the image quality is effectively assessed by comparing the query image with its pristine reference image.\nHowever, for the images in the wild, it is qu... | [
{
"id": "uzea9ZrUGH",
"initial_rating": 5,
"confidence": 5,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "This paper seeks to solve the no-reference image quality assessment issue with the help of distilled reference knowledge while eliminating the direct use of refer... | {
"rating": "3;3;5;5",
"rating_avg": 4,
"confidence": "5;4;5;5",
"confidence_avg": 4.75,
"soundness": "3;2;3;3",
"soundness_avg": 2.75,
"contribution": "2;3;2;2",
"contribution_avg": 2.25,
"presentation": "2;2;2;3",
"presentation_avg": 2.25
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:04.081627"
} | {
"id": "IhjpmJqsun",
"metareview": "This paper proposes a RKIQT by introducing the reference knowledge into the NR-IQA. Experimental results show the effectiveness of the proposed method. The major concerns of the reviewers include the limited novelty, insufficient evaluations and comparisons with state-of-the-a... | {
"decision": "Reject"
} |
wPMRwmytZe | 2410.05464v1 | Progressive distillation induces an implicit curriculum | {
"content": "## \n\n### Abstract\n\nKnowledge distillation leverages a teacher model to improve the training of a student model.\nA persistent challenge is that a better teacher does not always yield a better student, to which a common mitigation is to use additional supervision from several “intermediate” teachers.... | [
{
"id": "gxwbHHXl48",
"initial_rating": 6,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "Knowledge distillation is a widely used technique for training smaller \"student\" models by leveraging the knowledge captured by larger, pre-trained \"teacher\" ... | {
"rating": "6;6;6;8;8",
"rating_avg": 6.8,
"confidence": "3;3;3;4;3",
"confidence_avg": 3.2,
"soundness": "4;3;3;3;4",
"soundness_avg": 3.4,
"contribution": "3;3;3;4;3",
"contribution_avg": 3.2,
"presentation": "4;3;3;4;4",
"presentation_avg": 3.6
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Oral",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:04.082508"
} | {
"id": "vVsDF29qic",
"metareview": "The paper studies progressive distillation, where the student learns from a series of intermediate teacher checkpoints, as opposed to learning from a single, fully trained teacher. The authors propose that the improvement from progressive distillation stems from an \"implicit cu... | {
"decision": "Accept (Oral)"
} |
wPyTeUMRgh | 2410.02231v1 | SEAL: SEmantic-Augmented Imitation Learning via Language Model | {
"content": "## Abstract\n\nAbstract Hierarchical Imitation Learning (HIL) is a promising approach for tackling long-horizon decision-making tasks. While it is a challenging task due to the lack of detailed supervisory labels for sub-goal learning, and reliance on hundreds to thousands of expert demonstrations. In t... | [
{
"id": "0fPd1Hu4na",
"initial_rating": 3,
"confidence": 4,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "The authors present a hierarchical imitation learning method that tries to learn a latent subgoal representation.\nThe subgoal representation is learned from \n1.... | {
"rating": "3;3;5;6",
"rating_avg": 4.25,
"confidence": "4;4;5;4",
"confidence_avg": 4.25,
"soundness": "2;3;4;4",
"soundness_avg": 3.25,
"contribution": "2;2;3;3",
"contribution_avg": 2.5,
"presentation": "2;3;3;3",
"presentation_avg": 2.75
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:04.083046"
} | {
"id": "h51I2MfcTw",
"metareview": "The authors present a hierarchical imitation learning method that tries to learn a latent subgoal representation. The subgoal representation is learned from by (1) generating language subgoals from instructions using LLM prompting. (2) labeling the state from the expert demonstr... | {
"decision": "Reject"
} |
wQEdh2cgEk | 2410.11287v1 | Process Reward Model with Q-value Rankings | {
"content": "## Abstract\n\nAbstract Process Reward Modeling (PRM) is critical for complex reasoning and decision-making tasks where the accuracy of intermediate steps significantly influences the overall outcome. Existing PRM approaches, primarily framed as classification problems, employ cross-entropy loss to inde... | [
{
"id": "V2gzLZ0Ekr",
"initial_rating": 5,
"confidence": 4,
"soundness": 2,
"contribution": 3,
"presentation": 2,
"summary": "The paper introduces an algorithm to train a process verifier using a Q-value ranking objective. In particular, they split the intermediate steps in a generated t... | {
"rating": "3;8;8;8",
"rating_avg": 6.75,
"confidence": "4;3;3;2",
"confidence_avg": 3,
"soundness": "3;4;3;3",
"soundness_avg": 3.25,
"contribution": "2;3;4;3",
"contribution_avg": 3,
"presentation": "2;4;3;3",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:04.083885"
} | {
"id": "n0Vispr87X",
"metareview": "This paper presents a new approach to process reward modeling called PQM that focuses on optimizing Q-value rankings rather than treating the problem as a simple classification task. By doing so, PQM better captures the inherent relationships between steps in a reasoning process... | {
"decision": "Accept (Poster)"
} |
wQHyjIZ1SH | 2410.03535v1 | NRGBoost: Energy-Based Generative Boosted Trees | {
"content": "## Abstract\n\nAbstract Despite the rise to dominance of deep learning in unstructured data domains, tree-based methods such as Random Forests (RF) and Gradient Boosted Decision Trees (GBDT) are still the workhorses for handling discriminative tasks on tabular data. We explore generative extensions of t... | [
{
"id": "DIily5lQM5",
"initial_rating": 6,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "This paper explores generative extensions of tree-based models explicitly molding the data density for structured tabular data. Specifically, an energy-based gene... | {
"rating": "3;5;6;6",
"rating_avg": 5,
"confidence": "3;3;4;3",
"confidence_avg": 3.25,
"soundness": "1;3;2;3",
"soundness_avg": 2.25,
"contribution": "1;3;4;3",
"contribution_avg": 2.75,
"presentation": "1;4;2;3",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:04.084818"
} | {
"id": "b6tDJXhIWG",
"metareview": "The paper extends Gradient Boosted Decision Trees and Random Forests to generative modeling tasks. This paper went through significant changes during the review period, in a persistent effort from the authors to address comments from the reviewers, who eventually increased their... | {
"decision": "Accept (Poster)"
} |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.