paper_id string | arxiv_id string | title string | markdown dict | reviews list | scores dict | metadata dict | meta_review dict | decision dict |
|---|---|---|---|---|---|---|---|---|
FRzCIlkM7I | 2410.12593v1 | Expand and Compress: Exploring Tuning Principles for Continual Spatio-Temporal Graph Forecasting | {
"content": "## Abstract\n\nAbstract The widespread deployment of sensing devices leads to a surge in data for spatio-temporal forecasting applications such as traffic flow, air quality, and wind energy. Although spatio-temporal graph neural networks (STGNNs) have achieved success in modeling various static spatio-t... | [
{
"id": "2vJRm7E1B0",
"initial_rating": 3,
"confidence": 5,
"soundness": 2,
"contribution": 2,
"presentation": 3,
"summary": "This paper proposes EAC, a continuous spatio-temporal graph forecasting framework based on a continuous prompt parameter pool, aiming to address prediction challe... | {
"rating": "3;6;8;8",
"rating_avg": 6.25,
"confidence": "5;5;4;5",
"confidence_avg": 4.75,
"soundness": "2;3;3;4",
"soundness_avg": 3,
"contribution": "2;3;3;4",
"contribution_avg": 3,
"presentation": "3;3;3;4",
"presentation_avg": 3.25
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.404228"
} | {
"id": "aJl7KKxtCd",
"metareview": "This paper proposes a prompt tuning-based continuous forecasting method, EAC, following two fundamental tuning principles guided by empirical and theoretical analysis. Overall, the reviewers liked the motivation, rationale, method design, and evaluation results. Some reviewers... | {
"decision": "Accept (Poster)"
} |
FSjIrOm1vz | 2410.04343v1 | Inference Scaling for Long-Context Retrieval Augmented Generation | {
"content": "## Abstract\n\nAbstract The scaling of inference computation has unlocked the potential of long-context large language models (LLMs) across diverse settings. For knowledge-intensive tasks, the increased compute is often allocated to incorporate more external knowledge. However, without effectively utili... | [
{
"id": "7Q6KZbhhal",
"initial_rating": 8,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 4,
"summary": "This paper studies the inference scaling behaviors of two retrieval augmented generation (RAG) methods, demonstration-based RAG (DRAG) and iterative demonstration... | {
"rating": "6;8;8;8",
"rating_avg": 7.5,
"confidence": "4;2;3;3",
"confidence_avg": 3,
"soundness": "4;3;3;3",
"soundness_avg": 3.25,
"contribution": "3;3;3;3",
"contribution_avg": 3,
"presentation": "4;3;3;4",
"presentation_avg": 3.5
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Oral",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.404941"
} | {
"id": "qG4g1olN2u",
"metareview": "This paper presents a detailed investigation into the inference scaling of retrieval augmented generation (RAG) for long-context LLMs, exploring effective strategies for utilizing external knowledge beyond merely increasing its quantity. The authors focus on in-context learning ... | {
"decision": "Accept (Oral)"
} |
FSlfoBIctk | 2410.18533v1 | LOGO --- Long cOntext aliGnment via efficient preference Optimization | {
"content": "## Abstract\n\nAbstract Long-context models (LCMs) have shown great potential in processing long input sequences (even more than 100M tokens) conveniently and effectively.\nWith significant progress, recent research has pointed out that LCMs can accurately locate token-level salient information within t... | [
{
"id": "8LfQF7XJHt",
"initial_rating": 6,
"confidence": 5,
"soundness": 4,
"contribution": 4,
"presentation": 4,
"summary": "The paper introduces LOGO, a preference optimization training strategy to improve long-context alignment in language models. LOGO uses a reference-free preference... | {
"rating": "3;5;6;6",
"rating_avg": 5,
"confidence": "4;4;2;5",
"confidence_avg": 3.75,
"soundness": "3;2;2;4",
"soundness_avg": 2.75,
"contribution": "1;2;3;4",
"contribution_avg": 2.5,
"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": "2026-01-14T22:16:01.405710"
} | {
"id": "ryhzVCiSEv",
"metareview": "(a) The paper introduces LOGO, an efficient preference optimization method for long-context alignment using a reference-free approach and positional index synthesis. The model achieves competitive results with GPT-4 while requiring limited resources.\n\n(b) Strengths include com... | {
"decision": "Reject"
} |
FV6rPMwmuG | 2306.05300v2 | Anti-Correlated Noise in Epoch-Based Stochastic Gradient Descent: Implications for Weight Variances | {
"content": "## Abstract\n\nAbstract Stochastic gradient descent (SGD) has become a cornerstone of neural network optimization, yet the noise introduced by SGD is often assumed to be uncorrelated over time, despite the ubiquity of epoch-based training. In this work, we challenge this assumption and investigate the e... | [
{
"id": "6D6ZcIP6x9",
"initial_rating": 3,
"confidence": 4,
"soundness": 2,
"contribution": 1,
"presentation": 3,
"summary": "This paper studies the correlation of gradient noise in the later training phase of SGD. They show that SGD noise will be anti-correlated over time, assuming the ... | {
"rating": "3;5;5;6",
"rating_avg": 4.75,
"confidence": "4;3;3;4",
"confidence_avg": 3.5,
"soundness": "2;3;2;3",
"soundness_avg": 2.5,
"contribution": "1;2;3;3",
"contribution_avg": 2.25,
"presentation": "3;4;2;3",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:01.407581"
} | {
"id": "UaleIYWDX4",
"metareview": "This paper studies the anti-correlation of gradient noise for iterates of epoch-based (without-replacement) SGD and its implications to the variance of weight fluctuation at the end of training. Under the assumption that the noise of SGD is static (which holds e.g., when the noi... | {
"decision": "Reject"
} |
FVgizbs3o2 | 2307.00526v2 | TensorGPT: Efficient Compression of Large Language Models based on Tensor-Train Decomposition | {
"content": "## Abstract\n\nAbstract High-dimensional token embeddings underpin Large Language Models (LLMs), as they can capture subtle semantic information and significantly enhance the modelling of complex language patterns.\nHowever, this high dimensionality also introduces considerable model parameters and proh... | [
{
"id": "ec3ZSh7RRB",
"initial_rating": 3,
"confidence": 5,
"soundness": 2,
"contribution": 2,
"presentation": 3,
"summary": "This paper proposes using tensor-train decomposition to compress the token embedding matrix, aiming to reduce model size and accelerate inference, particularly in... | {
"rating": "3;3;3;6",
"rating_avg": 3.75,
"confidence": "5;5;5;3",
"confidence_avg": 4.5,
"soundness": "2;2;2;3",
"soundness_avg": 2.25,
"contribution": "2;1;2;3",
"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": "2026-01-14T22:16:01.408404"
} | {
"id": "JPCR4HxZDi",
"metareview": "This paper proposes TensorGPT, a method for compressing small language models (SLMs) using tensor-train decomposition (TTD) of the token embedding layer. The authors claim this approach is training-free and suitable for deploying SLMs on low-end devices. The method is evaluated ... | {
"decision": "Reject"
} |
FVuqJt3c4L | 2406.03044v2 | Population Transformer: Learning Population-level Representations of Neural Activity | {
"content": "## Abstract\n\nAbstract We present a self-supervised framework that learns population-level codes for arbitrary ensembles of neural recordings at scal. We address two key challenges in scaling models with neural time-series data: sparse and variable electrode distribution across subjects and datasets. T... | [
{
"id": "Ayq8tcDvS1",
"initial_rating": 5,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "The manuscript introduces the Population Transformer (PopT). This self-supervised framework tackles two key challenges in neural time-series analysis: sparse elec... | {
"rating": "3;5;5;5;8;8",
"rating_avg": 5.666666666666667,
"confidence": "3;5;2;3;3;2",
"confidence_avg": 3,
"soundness": "2;2;3;3;3;3",
"soundness_avg": 2.6666666666666665,
"contribution": "3;2;3;3;4;3",
"contribution_avg": 3,
"presentation": "1;1;1;3;2;3",
"presentation_avg": 1.8333333333333333
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Oral",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.409032"
} | {
"id": "AADYKBo2FW",
"metareview": "The paper introduces a self-supervised model called Population Transformer to model brain-wide neural activity sparsely and variably measured across subjects and datasets. Representations generated by this pre-trained model can then be used to perform downstream decoding tasks, ... | {
"decision": "Accept (Oral)"
} |
FXw0okNcOb | 2410.01949v1 | Discrete Copula Diffusion | {
"content": "## Abstract\n\nAbstract Discrete diffusion models have recently shown significant progress in modeling complex data, such as natural languages and DNA sequences. However, unlike diffusion models for continuous data, which can generate high-quality samples in just a few denoising steps, modern discrete d... | [
{
"id": "QpENasyqkb",
"initial_rating": 5,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 2,
"summary": "This paper studies the dependency issue of multivariables in each step of discrete diffusion model. The paper claims that discrete diffusion model requires many ... | {
"rating": "3;5;5;6",
"rating_avg": 4.75,
"confidence": "5;3;4;4",
"confidence_avg": 4,
"soundness": "3;2;3;3",
"soundness_avg": 2.75,
"contribution": "2;3;3;3",
"contribution_avg": 2.75,
"presentation": "2;3;2;3",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.409868"
} | {
"id": "kDR8JRmgHV",
"metareview": "This paper introduces Discrete Copula Diffusion (DCD), a framework to improve discrete diffusion models by addressing their inability to capture inter-variable dependencies at each denoising step. The method integrates a discrete diffusion model with an autoregressive copula mod... | {
"decision": "Accept (Poster)"
} |
FZv3kPHTtB | 2312.10300v2 | Shot2Story: A New Benchmark for Comprehensive Understanding of Multi-shot Videos | {
"content": "## Abstract\n\nAbstract A short clip of video may contain progression of multiple events and an interesting story line. A human need to capture both the event in every shot and associate them together to understand the story behind it. In this work, we present a new multi-shot video understanding benchm... | [
{
"id": "1uktail1L6",
"initial_rating": 5,
"confidence": 4,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "The paper introduces Shot2Story, a benchmark aimed at improving multi-shot video understanding through a rich set of annotations, including shot-level captions, d... | {
"rating": "5;5;6;6",
"rating_avg": 5.5,
"confidence": "4;4;2;4",
"confidence_avg": 3.5,
"soundness": "3;3;3;3",
"soundness_avg": 3,
"contribution": "3;2;3;3",
"contribution_avg": 2.75,
"presentation": "3;3;3;3",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.411073"
} | {
"id": "AZMIDZRsk8",
"metareview": "This paper introduces a multi-shot video understanding benchmark, providing captions for both visual signals and human narration. Based on this new benchmark, several tasks are evaluated to establish a foundation for future research.\nInitially, the submission received mixed fee... | {
"decision": "Accept (Poster)"
} |
FbQLFsBbTe | 2407.01445v3 | FastCLIP: A Suite of Optimization Techniques to Accelerate CLIP Training with Limited Resources | {
"content": "## Abstract\n\nAbstract Existing studies of training state-of-the-art Contrastive Language-Image Pretraining (CLIP) models on large-scale data involve hundreds of or even thousands of GPUs due to the requirement of a large batch size. However, such a large amount of resources is not accessible to most p... | [
{
"id": "PuVJBmrfn6",
"initial_rating": 3,
"confidence": 3,
"soundness": 2,
"contribution": 2,
"presentation": 2,
"summary": "This paper introduces FastCLIP, a distributed training framework designed to optimize CLIP model training using compositional optimization techniques, removing th... | {
"rating": "3;3;5",
"rating_avg": 3.6666666666666665,
"confidence": "4;3;3",
"confidence_avg": 3.3333333333333335,
"soundness": "2;2;3",
"soundness_avg": 2.3333333333333335,
"contribution": "2;2;2",
"contribution_avg": 2,
"presentation": "3;2;3",
"presentation_avg": 2.6666666666666665
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:01.411945"
} | {
"id": "zUe5irbSfR",
"metareview": "Thank you for your submission to ICLR. This paper introduces FastCLIP, a set of techniques and implementation for efficient training of CLIP in settings with limited resources. To carry this out, the authors developed a gradient reduction strategy to reduce communication require... | {
"decision": "Reject"
} |
FbbusgKmSW | 2406.04251v2 | Improving Gaussian Splatting with Localized Points Management | {
"content": "## Abstract\n\nAbstract Point management is a critical component in optimizing 3D Gaussian Splatting (3DGS) models, as the point initiation (e.g., via structure from motion) is distributionally inappropriate.\nTypically, the Adaptive Density Control (ADC) algorithm is applied, leveraging view-averaged g... | [
{
"id": "1Wg5o8JGby",
"initial_rating": 5,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "The paper introduces a point management method for 3D Gaussian Splatting, designed to improve point densification and correction. This approach leverages renderin... | {
"rating": "5;5;6",
"rating_avg": 5.333333333333333,
"confidence": "4;4;4",
"confidence_avg": 4,
"soundness": "2;3;3",
"soundness_avg": 2.6666666666666665,
"contribution": "1;3;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": "2026-01-14T22:16:01.412739"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
Fg0eo2AkST | 2402.04236v2 | CogCoM: A Visual Language Model with Chain-of-Manipulations Reasoning | {
"content": "## Abstract\n\nAbstract Vision-Language Models (VLMs) have demonstrated their broad effectiveness thanks to extensive training in aligning visual instructions to responses.\nHowever, such training of conclusive alignment leads models to ignore essential visual reasoning, further resulting in failures in... | [
{
"id": "HYKPBsF6dK",
"initial_rating": 6,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "The paper focuses on visual reasoning problems using large vision language models. The paper focuses on the problem caused by the first alignment training stage i... | {
"rating": "5;5;6;6",
"rating_avg": 5.5,
"confidence": "4;4;4;4",
"confidence_avg": 4,
"soundness": "3;3;4;3",
"soundness_avg": 3.25,
"contribution": "2;2;3;3",
"contribution_avg": 2.5,
"presentation": "3;1;4;3",
"presentation_avg": 2.75
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.413339"
} | {
"id": "eeH3ssCBnZ",
"metareview": "This paper introduces a mechanism called “Chain of Manipulations” inspired by human cognition, which enables VLM to solve problems step by step using evidence. With this mechanism, the model can extract unique operations and solve visual problems. In addition, it is also possibl... | {
"decision": "Accept (Poster)"
} |
FgirWC5TJ6 | 2410.04421v2 | Disentangling Regional Primitives for Image Generation | {
"content": "## Abstract\n\nAbstract This paper presents a method to explain the internal representation structure of a neural network for image generation. Specifically, our method disentangles primitive feature components from the intermediate-layer feature of the neural network, which ensures that each feature co... | [
{
"id": "XbCf1dGU6D",
"initial_rating": 5,
"confidence": 3,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "This paper proposes a method to interprete the GAN model by disentangling regisonal primitives for image generation. Following the similar AND definition proposed... | {
"rating": "3;5;5;5",
"rating_avg": 4.5,
"confidence": "3;4;3;3",
"confidence_avg": 3.25,
"soundness": "3;2;2;3",
"soundness_avg": 2.5,
"contribution": "2;2;2;2",
"contribution_avg": 2,
"presentation": "1;3;1;3",
"presentation_avg": 2
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:01.413995"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
FhBT596F1X | 2410.07972v1 | Learning Equivariant Non-Local Electron Density Functionals | {
"content": "## Abstract\n\nAbstract The accuracy of density functional theory hinges on the approximation of non-local contributions to the exchange-correlation (XC) functional.\nTo date, machine-learned and human-designed approximations suffer from insufficient accuracy, limited scalability, or dependence on costl... | [
{
"id": "dCHgdfnaGX",
"initial_rating": 8,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "This paper proposes to learn weighting coefficients for every point in 3D space to re-weight meta-GGA functional.",
"strengths": "- The proposed method is bas... | {
"rating": "5;6;6;8",
"rating_avg": 6.25,
"confidence": "3;4;3;3",
"confidence_avg": 3.25,
"soundness": "2;4;3;3",
"soundness_avg": 3,
"contribution": "2;3;2;3",
"contribution_avg": 2.5,
"presentation": "3;3;3;3",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Spotlight",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.414703"
} | {
"id": "PIwJRRbHSL",
"metareview": "The submission proposes a new way of modeling and learning the exchange-correlation (XC) functional for more accurate Kohn-Sham density functional theory (DFT) calculations. The work models the electron density as atom-centered equivariant coefficients, which is processed by equ... | {
"decision": "Accept (Spotlight)"
} |
FhhH14jso4 | 2407.12068v2 | Learning on Graphs with Large Language Models (LLMs): A Deep Dive into Model Robustness | {
"content": "## Abstract\n\nAbstract Large Language Models (LLMs) have demonstrated remarkable performance across various natural language processing tasks. Recently, several LLMs-based pipelines have been developed to enhance learning on graphs with text attributes, showcasing promising performance. However, graphs... | [
{
"id": "ayYfx95rxu",
"initial_rating": 3,
"confidence": 3,
"soundness": 2,
"contribution": 2,
"presentation": 2,
"summary": "The paper investigates the robustness of graph machine learning methods with LLMs (Graph-LLMs) against adversarial attacks. More specifically, the investigation i... | {
"rating": "3;3;5;5",
"rating_avg": 4,
"confidence": "4;3;5;4",
"confidence_avg": 4,
"soundness": "2;2;3;2",
"soundness_avg": 2.25,
"contribution": "2;2;2;2",
"contribution_avg": 2,
"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": "2026-01-14T22:16:01.415412"
} | {
"id": "H5roqSMzfh",
"metareview": "This paper focuses on providing empirical evaluation of LLMs over graphs in terms of its impact on the robustness. The reviewers are concerned about the limited attacks studied in the experiments, the inconsistent metrics used across different experiments, as well as the limited... | {
"decision": "Reject"
} |
Fk3eod9aaD | 2410.08258v1 | In Search of Forgotten Domain Generalization | {
"content": "## Abstract\n\nAbstract Out-of-Domain (OOD) generalization is the ability of a model trained on one or more domains to generalize to unseen domains. In the ImageNet era of computer vision, evaluation sets for measuring a model’s OOD performance were designed to be strictly OOD with respect to style. How... | [
{
"id": "IAoAEsRGBY",
"initial_rating": 8,
"confidence": 4,
"soundness": 4,
"contribution": 4,
"presentation": 4,
"summary": "This paper investigates if the recent purported gains in generalization reported for Large scale vision language models come from true generalization beyond the d... | {
"rating": "3;6;8;8",
"rating_avg": 6.25,
"confidence": "3;4;4;4",
"confidence_avg": 3.75,
"soundness": "2;3;3;4",
"soundness_avg": 3,
"contribution": "2;2;3;4",
"contribution_avg": 2.75,
"presentation": "3;3;3;4",
"presentation_avg": 3.25
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Spotlight",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.415918"
} | {
"id": "zBt2hl4pqG",
"metareview": "The paper addresses a critical and timely issue of out-of-distribution (OOD) generalization in the era of foundation models.\nIt challenges the common assumption that models like CLIP inherently generalize OOD, providing evidence that their strong performance often stems from do... | {
"decision": "Accept (Spotlight)"
} |
FmoInsWCkp | 2410.14426v1 | Predicting Time-Varying Flux and Balance in Metabolic Systems using Structured Neural ODE Processes | {
"content": "## Abstract\n\nAbstract We develop a novel data-driven framework as an alternative to dynamic flux balance analysis, bypassing the demand for deep domain knowledge and manual efforts to formulate the optimization problem. The proposed framework is end-to-end, which trains a structured neural ODE process... | [
{
"id": "rQKOHCXdcO",
"initial_rating": 5,
"confidence": 3,
"soundness": 2,
"contribution": 2,
"presentation": 2,
"summary": "This paper addresses the challenging problem of predicting metabolic flux and metabolic balance from single-cell data. The authors formulate this prediction probl... | {
"rating": "3;3;3;5",
"rating_avg": 3.5,
"confidence": "2;4;3;3",
"confidence_avg": 3,
"soundness": "2;1;2;2",
"soundness_avg": 1.75,
"contribution": "1;1;1;2",
"contribution_avg": 1.25,
"presentation": "2;2;2;2",
"presentation_avg": 2
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:01.416535"
} | {
"id": "nMjNpg56qN",
"metareview": "The reviewers and the authors engaged in a good discussion. Although the reviewers appreciate the problem tackled in the paper (neural ODEs for metabolic networks), the reviewers were not convinced that the paper is ready for ICLR.\n\nThe authors are encouraged to continue their... | {
"decision": "Reject"
} |
FoF5RaA3ug | 2405.14736v1 | GIFT: Unlocking Full Potential of Labels in Distilled Dataset at Near-zero Cost | {
"content": "## Abstract\n\nAbstract Recent advancements in dataset distillation have demonstrated the significant benefits of employing soft labels generated by pre-trained teacher models.\nIn this paper, we introduce a novel perspective by emphasizing the full utilization of labels.\nWe first conduct a comprehensi... | [
{
"id": "1dLmbdivhx",
"initial_rating": 6,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "This paper conducts a comprehensive comparison of various loss functions for soft label utilization in DD and find that the model trained on the synthetic dataset... | {
"rating": "5;6;6;6",
"rating_avg": 5.75,
"confidence": "4;5;4;4",
"confidence_avg": 4.25,
"soundness": "3;3;3;3",
"soundness_avg": 3,
"contribution": "2;2;3;3",
"contribution_avg": 2.5,
"presentation": "3;3;4;3",
"presentation_avg": 3.25
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.417217"
} | {
"id": "qk5Y88RhWx",
"metareview": "This work comprehensive evaluates several loss functions for dataset distillation and reveals the high sensitivity of model trained on the\nsynthetic dataset to the choice of loss function. leveraging this finding, the authors introduce a new method called GIFT, which is a simpl... | {
"decision": "Accept (Poster)"
} |
FoUpv84hMw | 2405.18795v1 | Federated $Q$-Learning with Reference-Advantage Decomposition: Almost Optimal Regret and Logarithmic Communication Cost | {
"content": "## Abstract\n\nAbstract In this paper, we consider model-free federated reinforcement learning for tabular episodic Markov decision processes. Under the coordination of a central server, multiple agents collaboratively explore the environment and learn an optimal policy without sharing their raw data. D... | [
{
"id": "zUbkDxZQqT",
"initial_rating": 6,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "The paper addresses the problem of federated reinforcement learning (RL) in a model-free, tabular episodic Markov Decision Process setting. The FedQ-Advantage al... | {
"rating": "3;5;5;6;6;8",
"rating_avg": 5.5,
"confidence": "4;3;3;3;3;4",
"confidence_avg": 3.3333333333333335,
"soundness": "3;3;3;3;3;3",
"soundness_avg": 3,
"contribution": "2;3;3;3;3;2",
"contribution_avg": 2.6666666666666665,
"presentation": "1;2;2;2;3;3",
"presentation_avg": 2.166666666666666... | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.418295"
} | {
"id": "GWW2Oyhf4u",
"metareview": "The authors proposed the FedQ-Advantage Algorithm for federated reinforcement learning, enabling multiple agents to learn collaboratively through a central server. The algorithm improves upon existing methods by: (1) reducing communication between agents and the server by sharin... | {
"decision": "Accept (Poster)"
} |
FowFLhUTgO | 2410.10382v1 | V2M: Visual 2-Dimensional Mamba for Image Representation Learning | {
"content": "## Abstract\n\nAbstract Mamba has garnered widespread attention due to its flexible design and efficient hardware performance to process 1D sequences based on the state space model (SSM).\nRecent studies have attempted to apply Mamba to the visual domain by flattening 2D images into patches and then reg... | [
{
"id": "90znc37Gwa",
"initial_rating": 6,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "The paper proposes a 2D state-space model (SSM) for visual representation learning, named V2M. Unlike previous Mamba-based visual representation methods that perf... | {
"rating": "3;5;6;8",
"rating_avg": 5.5,
"confidence": "3;5;4;5",
"confidence_avg": 4.25,
"soundness": "3;3;3;3",
"soundness_avg": 3,
"contribution": "2;3;3;3",
"contribution_avg": 2.75,
"presentation": "2;3;3;2",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:01.419287"
} | {
"id": "yU1aCWDJNH",
"metareview": "This paper proposes a Mamba-based backbone for vision tasks, using a 2D SSM formulation to effectively incorporate 2D locality priors into vision data. Reviewers generally find the motivation for this approach intuitive and interesting; however, they raised significant concerns ... | {
"decision": "Reject"
} |
FpiCLJrSW8 | 2404.18870v1 | More RLHF, More Trust? On The Impact of Preference Alignment On Trustworthiness | {
"content": "## Abstract\n\nAbstract The surge in Large Language Models (LLMs) development has led to improved performance on cognitive tasks as well as an urgent need to align these models with human values in order to safely exploit their power. Despite the effectiveness of preference learning algorithms like Rein... | [
{
"id": "6SX5ZQyxsw",
"initial_rating": 8,
"confidence": 3,
"soundness": 4,
"contribution": 3,
"presentation": 4,
"summary": "The paper evaluates how Reinforcement Learning from Human Feedback (RLHF) impacts Large Language Model (LLM) trustworthiness. It assesses five dimensions: toxicit... | {
"rating": "5;6;6;8",
"rating_avg": 6.25,
"confidence": "4;3;3;3",
"confidence_avg": 3.25,
"soundness": "2;3;3;4",
"soundness_avg": 3,
"contribution": "2;3;3;3",
"contribution_avg": 2.75,
"presentation": "1;4;3;4",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Oral",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.419811"
} | {
"id": "yFUM8FhErp",
"metareview": "This paper evaluates four language models Pythia 1.4B, 2.8B and 6.9B as well as Llama-&B on five axes of “trustworthiness” : Toxicity, Stereotypical bias, Machine Ethics, Truthfulness and Privacy. The main contribution is that optimizing these LLMs for human preferences\ndoes n... | {
"decision": "Accept (Oral)"
} |
FpxKYYk6V5 | 2406.11501v2 | Teleporter Theory: A General and Simple Approach for Modeling Cross-World Counterfactual Causality | {
"content": "## Abstract\n\nAbstract Leveraging the development of structural causal model (SCM), researchers can establish graphical models for exploring the causal mechanisms behind machine learning techniques. As the complexity of machine learning applications rises, single-world interventionism causal analysis e... | [
{
"id": "8W8qRUWMI9",
"initial_rating": 3,
"confidence": 4,
"soundness": 3,
"contribution": 1,
"presentation": 3,
"summary": "This paper points out the issues with existing frameworks for counterfactual representations, such as twin networks or SWIG. The authors propose the teleporter th... | {
"rating": "3;3;6",
"rating_avg": 4,
"confidence": "3;4;3",
"confidence_avg": 3.3333333333333335,
"soundness": "2;3;2",
"soundness_avg": 2.3333333333333335,
"contribution": "2;1;4",
"contribution_avg": 2.3333333333333335,
"presentation": "2;3;3",
"presentation_avg": 2.6666666666666665
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:01.420517"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
FrjTgprk3V | 2407.20199v2 | Emergence in non-neural models: grokking modular arithmetic via average gradient outer product | {
"content": "## Abstract\n\nAbstract Neural networks trained to solve modular arithmetic tasks exhibit grokking , a phenomenon where the test accuracy starts improving long after the model achieves 100 % percent 100 100\\% 100 % training accuracy in the training process. It is often taken as an example of “emergence... | [
{
"id": "SJBfapxIJ5",
"initial_rating": 8,
"confidence": 3,
"soundness": 4,
"contribution": 3,
"presentation": 4,
"summary": "This paper studies the phenomenon of grokking in non-neural models, specifically Recursive Feature Machines (RFMs). In the setting they study (modular arithmetic)... | {
"rating": "5;5;8",
"rating_avg": 6,
"confidence": "4;4;3",
"confidence_avg": 3.6666666666666665,
"soundness": "2;3;4",
"soundness_avg": 3,
"contribution": "2;2;3",
"contribution_avg": 2.3333333333333335,
"presentation": "3;4;4",
"presentation_avg": 3.6666666666666665
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:01.421457"
} | {
"id": "0SuXIKZWKi",
"metareview": "The authors present an example of grokking on modular arithmetic in recursive feature machines (RFMs), which are neither neural networks nor trained by gradient descent. While grokking has arguably been shown in linear models (https://arxiv.org/abs/2310.16441) it has not been sh... | {
"decision": "Reject"
} |
Frok9AItud | 2404.10148v2 | Node Similarities under Random Projections: Limits and Pathological Cases | {
"content": "## Abstract\n\nAbstract Random Projections have been widely used to generate embeddings for various graph learning tasks due to their computational efficiency. The majority of applications have been justified through the Johnson-Lindenstrauss Lemma. In this paper, we take a step further and investigate ... | [
{
"id": "YYZgNLi50w",
"initial_rating": 8,
"confidence": 3,
"soundness": 3,
"contribution": 4,
"presentation": 4,
"summary": "The paper focuses on the practice of using random projections (RP) to learn graph embeddings. In particular, they motivate the advantage of using Cosine Similarit... | {
"rating": "3;3;6;6;8",
"rating_avg": 5.2,
"confidence": "5;4;2;4;3",
"confidence_avg": 3.6,
"soundness": "2;3;3;3;3",
"soundness_avg": 2.8,
"contribution": "2;2;3;3;4",
"contribution_avg": 2.8,
"presentation": "1;2;3;3;4",
"presentation_avg": 2.6
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.422655"
} | {
"id": "0L4Wo1LTFl",
"metareview": "This paper studies random projection methods for graph embedding: in particular embeddings produced by AR or TR where R is a Johnson-Lindenstrauss random projection matrix and A and T are the graph adjacency matrices and transition matrices in particular. They argue that due to ... | {
"decision": "Accept (Poster)"
} |
Fs9EabmQrJ | 2410.02223v2 | EmbedLLM: Learning Compact Representations of Large Language Models | {
"content": "## Abstract\n\nAbstract With hundreds of thousands of language models available on Huggingface today, efficiently evaluating and utilizing these models across various downstream tasks has become increasingly critical. Many existing methods repeatedly learn task-specific representations of Large Language... | [
{
"id": "mO9UDpUGtq",
"initial_rating": 5,
"confidence": 3,
"soundness": 2,
"contribution": 3,
"presentation": 2,
"summary": "This paper proposes a novel framework for generating compact vector representations of LLMs to enhance model routing, task efficiency, and performance forecasting... | {
"rating": "5;6;6",
"rating_avg": 5.666666666666667,
"confidence": "3;4;4",
"confidence_avg": 3.6666666666666665,
"soundness": "2;3;3",
"soundness_avg": 2.6666666666666665,
"contribution": "3;3;4",
"contribution_avg": 3.3333333333333335,
"presentation": "2;3;2",
"presentation_avg": 2.33333333333333... | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Spotlight",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.423753"
} | {
"id": "Iq0ftIfdIu",
"metareview": "This paper presents an approach for learning a representation of the language model which helps it produce an embedding for a more compact representation of the language model. This allows the embedding to be used in three different tasks of interest: model routing, accuracy pre... | {
"decision": "Accept (Spotlight)"
} |
Fty0wTcemV | 2411.04425v2 | DELIFT: Data Efficient Language model Instruction Fine-Tuning | {
"content": "## Abstract\n\nAbstract Fine-tuning large language models (LLMs) is essential for enhancing their performance on specific tasks but is often resource-intensive due to redundant or uninformative data. To address this inefficiency, we introduce DELIFT (Data Efficient Language model Instruction Fine-Tuning... | [
{
"id": "tq4i1OA1f9",
"initial_rating": 5,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "This paper presents DELIFT, Data Efficient Language model Instruction Fine-Tuning, a method to select and prune fine-tuning data for three stages - initial instru... | {
"rating": "5;5;6",
"rating_avg": 5.333333333333333,
"confidence": "3;4;3",
"confidence_avg": 3.3333333333333335,
"soundness": "2;3;3",
"soundness_avg": 2.6666666666666665,
"contribution": "3;3;3",
"contribution_avg": 3,
"presentation": "3;3;2",
"presentation_avg": 2.6666666666666665
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.424384"
} | {
"id": "sUwvvq5hkZ",
"metareview": "## Summary: \nThe paper introduces DELIFT (Data Efficient Language model Instruction Fine-Tuning), an algorithm designed to optimize data selection efficiently across three crucial stages of fine-tuning large language models (LLMs). DELIFT utilizes a pairwise utility metric, whi... | {
"decision": "Accept (Poster)"
} |
FvBTy5Dz9C | 2409.02322v1 | TimeDiT: General-purpose Diffusion Transformers for Time Series Foundation Model | {
"content": "## Abstract\n\nAbstract With recent advances in building foundation models for texts and video data, there is a surge of interest in foundation models for time series. A family of models have been developed, utilizing a temporal auto-regressive generative Transformer architecture, whose effectiveness ha... | [
{
"id": "LZb7FjuItJ",
"initial_rating": 3,
"confidence": 2,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "The proposed work presents a method for training a diffusion model for time-series based on masked reconstruction. This model models the distribution of masked ti... | {
"rating": "3;3;6;8",
"rating_avg": 5,
"confidence": "4;2;4;4",
"confidence_avg": 3.5,
"soundness": "1;3;3;4",
"soundness_avg": 2.75,
"contribution": "2;2;3;3",
"contribution_avg": 2.5,
"presentation": "2;3;2;3",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:01.425161"
} | {
"id": "owGH56n9On",
"metareview": "This paper introduces Time Diffusion Transformer (TimeDiT), a diffusion-based generative model for time series. The authors provide experiments on a variety of tasks, such as forecasting, imputation, and anomaly detection and claim that their work provides a “general foundation ... | {
"decision": "Reject"
} |
FvIASa0tau | 2410.10790v1 | Sitcom-Crafter: A Plot-Driven Human Motion Generation System in 3D Scenes | {
"content": "## Abstract\n\nAbstract Recent advancements in human motion synthesis have focused on specific types of motions, such as human-scene interaction, locomotion or human-human interaction, however, there is a lack of a unified system capable of generating a diverse combination of motion types. In response, ... | [
{
"id": "jdmCCHs1ar",
"initial_rating": 3,
"confidence": 3,
"soundness": 2,
"contribution": 2,
"presentation": 2,
"summary": "The paper introduces \"Sitcom-Crafter,\" a system designed for generating two-person human motions in 3D environments, targeting animation and gaming industries. ... | {
"rating": "3;6;8",
"rating_avg": 5.666666666666667,
"confidence": "3;5;4",
"confidence_avg": 4,
"soundness": "2;2;3",
"soundness_avg": 2.3333333333333335,
"contribution": "2;3;4",
"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": "2026-01-14T22:16:01.426156"
} | {
"id": "0fN9V252ZT",
"metareview": "This paper, titled Sitcom-Crafter, presents a framework that unifies multiple motion synthesis tasks—human-human interaction, human-scene interaction, and locomotion—into a single pipeline driven by a plot-like textual description. The authors incorporate and adapt several exist... | {
"decision": "Accept (Poster)"
} |
Fvfs0HPuKl | 2403.00946v2 | fine-tuning with very large dropout | {
"content": "## Abstract\n\nAbstract It is impossible today to pretend that the practice of machine learning is compatible with the idea that training and testing data follow the same distribution. Several authors have recently used ensemble techniques to show how scenarios involving multiple data distributions are ... | [
{
"id": "wkqC0XgJ3i",
"initial_rating": 5,
"confidence": 4,
"soundness": 2,
"contribution": 2,
"presentation": 3,
"summary": "In the paper, the authors propose a novel, effective fine-tuning method to fix reduced performance on out-of-distribution (OOD) data. In essence, they begin with ... | {
"rating": "3;5;5;5;5",
"rating_avg": 4.6,
"confidence": "5;4;4;3;4",
"confidence_avg": 4,
"soundness": "2;3;2;3;2",
"soundness_avg": 2.4,
"contribution": "1;2;2;2;2",
"contribution_avg": 1.8,
"presentation": "2;3;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": "2026-01-14T22:16:01.427083"
} | {
"id": "SY9anWu4qh",
"metareview": "This paper shows that fine-tuning large pre-trained models with a very high dropout rate applied to the penultimate layer improves out-of-distribution (OOD) generalization. The authors explain it as providing \"richer\" feature representations beneficial for OOD tasks. The appro... | {
"decision": "Reject"
} |
FyJaV0TVF2 | 2410.12850v1 | RecurFormer: Not All Transformer Heads Need Self-Attention | {
"content": "## Abstract\n\nAbstract Transformer-based large language models (LLMs) excel in modeling complex language patterns but face significant computational costs during inference, especially with long inputs due to the attention mechanism’s memory overhead. We observe that certain attention heads exhibit a di... | [
{
"id": "egOFV5ebrr",
"initial_rating": 3,
"confidence": 4,
"soundness": 2,
"contribution": 2,
"presentation": 2,
"summary": "This paper proposed RecurFormer, which replaces parts of attention heads with linear recurrent neural networks (RNNs) to focus on local dependencies, reducing mem... | {
"rating": "3;3;3;5",
"rating_avg": 3.5,
"confidence": "4;4;4;4",
"confidence_avg": 4,
"soundness": "1;3;2;2",
"soundness_avg": 2,
"contribution": "2;3;2;2",
"contribution_avg": 2.25,
"presentation": "3;3;2;2",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:01.427959"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
FyMjfDQ9RO | 2410.07168v1 | Sylber: Syllabic Embedding Representation of Speech from Raw Audio | {
"content": "## Abstract\n\nAbstract † † footnotetext: Correspondence to: Cheol Jun Cho <cheoljun@berkeley.edu> , Gopala K. Anumanchipalli <gopala@berkeley.edu> Syllables are compositional units of spoken language that play a crucial role in human speech perception and production.\nHowever, current neural speech rep... | [
{
"id": "aJg7O47CfR",
"initial_rating": 6,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "The paper presents an innovative self-supervised learning (SSL) method that converts speech into syllable-based embeddings. The authors employ a range of evaluati... | {
"rating": "3;5;6;8",
"rating_avg": 5.5,
"confidence": "4;4;4;5",
"confidence_avg": 4.25,
"soundness": "1;3;3;4",
"soundness_avg": 2.75,
"contribution": "2;2;3;4",
"contribution_avg": 2.75,
"presentation": "3;3;3;4",
"presentation_avg": 3.25
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.428599"
} | {
"id": "ofte3x22Aa",
"metareview": "The paper proposes a self-supervised approach to learning speech representations. The approach involves distilling from a running average, and an unsupervised segmentation. The syllable information is more prominent compared to other approaches.\n\nThis is one of those rare case... | {
"decision": "Accept (Poster)"
} |
G19piTjVYA | 2410.08787v1 | Efficient Differentiable Discovery of Causal Order | {
"content": "## Abstract\n\nAbstract In the algorithm Intersort, Chevalley et al. ( 2024 ) proposed a score-based method to discover the causal order of variables in a Directed Acyclic Graph (DAG) model, leveraging interventional data to outperform existing methods. However, as a score-based method over the permutah... | [
{
"id": "Cxf2iPkTUl",
"initial_rating": 5,
"confidence": 3,
"soundness": 2,
"contribution": 3,
"presentation": 2,
"summary": "This paper focuses on a differentiable version of the Intersort score, a score for identifying the topological ordering of a causal graph via interventional data ... | {
"rating": "3;3;5;5",
"rating_avg": 4,
"confidence": "3;4;4;3",
"confidence_avg": 3.5,
"soundness": "2;2;2;2",
"soundness_avg": 2,
"contribution": "1;1;3;3",
"contribution_avg": 2,
"presentation": "3;2;1;2",
"presentation_avg": 2
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:01.430166"
} | {
"id": "vrJNeROJW6",
"metareview": "This paper introduces a differentiable version of the Intersort algorithm, and proposes CausalDisco, a causal discovery method using DiffIntersort for regularization. Reviewers found the scalability and differentiability promising but raised concerns about limited theoretical an... | {
"decision": "Reject"
} |
G1n50BMqzm | 2410.05586v2 | TeaserGen: Generating Teasers for Long Documentaries | {
"content": "## Abstract\n\nAbstract Teasers are an effective tool for promoting content in entertainment, commercial and educational fields. However, creating an effective teaser for long videos is challenging for it requires long-range multimodal modeling on the input videos, while necessitating maintaining audiov... | [
{
"id": "Ey3Q6gteW4",
"initial_rating": 6,
"confidence": 3,
"soundness": 2,
"contribution": 3,
"presentation": 3,
"summary": "This paper introduces TeaserGen, a method for generating teasers for long videos. To address the lack of suitable datasets, it presents the DocumentaryNet dataset... | {
"rating": "3;3;5;6",
"rating_avg": 4.25,
"confidence": "4;5;3;3",
"confidence_avg": 3.75,
"soundness": "2;3;2;2",
"soundness_avg": 2.25,
"contribution": "2;3;2;3",
"contribution_avg": 2.5,
"presentation": "2;3;3;3",
"presentation_avg": 2.75
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.430937"
} | {
"id": "PE3qcHnfy4",
"metareview": "The paper introduces a novel method for generating teaser clips, accompanied by a new benchmark. It received mixed reviews: two negative and two positive. The concerns raised were mostly minor, focusing on engineering aspects and stemming from misunderstandings. Although the aut... | {
"decision": "Accept (Poster)"
} |
G2BiEoB77Z | 2312.06343v1 | RankMatch: A Novel Approach to Semi-Supervised Label Distribution Learning Leveraging Inter-label Correlations | {
"content": "## Abstract\n\nAbstract This paper introduces RankMatch, an innovative approach for Semi-Supervised Label Distribution Learning (SSLDL). Addressing the challenge of limited labeled data, RankMatch effectively utilizes a small number of labeled examples in conjunction with a larger quantity of unlabeled ... | [
{
"id": "nYTXMfQNti",
"initial_rating": 5,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "The paper proposes a pairwise relevance ranking loss which captures inter-label correlations to tackle Semi-supervised Label Distribution Learning.",
"strengt... | {
"rating": "3;3;5;5",
"rating_avg": 4,
"confidence": "4;3;4;4",
"confidence_avg": 3.75,
"soundness": "2;2;3;3",
"soundness_avg": 2.5,
"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": "2026-01-14T22:16:01.431651"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
G2p8TLuJgy | 2406.00083v2 | BadRAG: Identifying Vulnerabilities in Retrieval Augmented Generation of Large Language Models | {
"content": "## Abstract\n\nAbstract Large Language Models (LLMs) are constrained by outdated information and a tendency to generate incorrect data, commonly referred to as \"hallucinations.\" Retrieval-Augmented Generation (RAG) addresses these limitations by combining the strengths of retrieval-based methods and g... | [
{
"id": "bnsLq85YTZ",
"initial_rating": 3,
"confidence": 5,
"soundness": 2,
"contribution": 1,
"presentation": 3,
"summary": "This paper proposes BadRAG as an attack on RAG. It is identified that poisoning several customized content passages could achieve a retrieval backdoor, where the ... | {
"rating": "3;3;3;5",
"rating_avg": 3.5,
"confidence": "4;5;5;5",
"confidence_avg": 4.75,
"soundness": "2;3;2;3",
"soundness_avg": 2.5,
"contribution": "1;2;1;3",
"contribution_avg": 1.75,
"presentation": "3;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": "2026-01-14T22:16:01.432259"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
G328D1xt4W | 2410.13643v1 | Fine-Tuning Discrete Diffusion Models via Reward Optimization with Applications to DNA and Protein Design | {
"content": "## Abstract\n\nAbstract Recent studies have demonstrated the strong empirical performance of diffusion models on discrete sequences (i.e., discrete diffusion models) across domains from natural language to biological sequence generation. For example, in the protein inverse folding task, where the goal i... | [
{
"id": "gB2hwvcNOj",
"initial_rating": 5,
"confidence": 2,
"soundness": 2,
"contribution": 3,
"presentation": 3,
"summary": "This paper presents a new fine-tuning method for discrete diffusion models. The objective is to generate high-reward yet natural sequences of DNA or RNA, where na... | {
"rating": "3;3;5;6",
"rating_avg": 4.25,
"confidence": "3;4;2;3",
"confidence_avg": 3,
"soundness": "3;2;2;3",
"soundness_avg": 2.5,
"contribution": "3;1;3;2",
"contribution_avg": 2.25,
"presentation": "2;2;3;3",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.433012"
} | {
"id": "5pgryjRLmK",
"metareview": "This paper introduces a novel fine-tuning approach for discrete diffusion models, aiming to generate sequences of DNA or RNA that achieve high rewards while maintaining naturalness. Naturalness is quantified using the KL divergence from the prior distribution represented by the ... | {
"decision": "Accept (Poster)"
} |
G3aXjVAJjU | 2410.22315v1 | Natural Language Inference Improves Compositionality in Vision-Language Models | {
"content": "## Abstract\n\nAbstract Compositional reasoning in Vision-Language Models (VLMs) remains challenging as these models often struggle to relate objects, attributes, and spatial relationships. Recent methods aim to address these limitations by relying on the semantics of the textual description, using Larg... | [
{
"id": "eRfBblHE11",
"initial_rating": 6,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "The paper presents Caption Expansion with Contradictions and Entailments (CECE), a new approach to improve compositionality in the vision-language models by trans... | {
"rating": "6;6;6;6",
"rating_avg": 6,
"confidence": "3;4;3;3",
"confidence_avg": 3.25,
"soundness": "3;3;3;3",
"soundness_avg": 3,
"contribution": "3;3;3;3",
"contribution_avg": 3,
"presentation": "3;3;3;3",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.433824"
} | {
"id": "gJGktOtEpS",
"metareview": "CECE improves compositional reasoning in vision-language models (VLMs) by using a large language model (LLM) to generate entailments and contradictions of captions. These expanded captions provide more diverse cues for better visual-textual alignment, leading to improved perform... | {
"decision": "Accept (Poster)"
} |
G5sPv4KSjR | 2408.16286v2 | Near-Optimal Policy Identification in Robust Constrained Markov Decision Processes via Epigraph Form | {
"content": "## Abstract\n\nAbstract Designing a safe policy for uncertain environments is crucial in real-world control applications.\nHowever, this challenge remains inadequately addressed within the Markov decision process (MDP) framework.\nThis paper presents the first algorithm capable of identifying a near-opt... | [
{
"id": "mMtY4ChaLp",
"initial_rating": 6,
"confidence": 3,
"soundness": 4,
"contribution": 3,
"presentation": 3,
"summary": "This paper proposes the EpiRC-PGS algorithm for solving the robust constrained Markov Decision Process (RCMDP) problem and provides a theoretical proof for identi... | {
"rating": "3;6;6;8",
"rating_avg": 5.75,
"confidence": "4;4;4;4",
"confidence_avg": 4,
"soundness": "2;4;3;3",
"soundness_avg": 3,
"contribution": "1;3;3;3",
"contribution_avg": 2.5,
"presentation": "1;3;3;3",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.435007"
} | {
"id": "PFZV6t6thZ",
"metareview": "This paper studies the problem of identifying near-optimal policies in Robust Constrained Markov Decision Processes (RCMDPs) using a novel approach based on the epigraph formulation. It achieves theoretical guarantees with convergence to an $\\epsilon$-optimal policy, addressing... | {
"decision": "Accept (Poster)"
} |
G6aJyS0ZV0 | 2405.18416v2 | 3D StreetUnveiler with Semantic-aware 2DGS - a simple baseline | {
"content": "## Abstract\n\nAbstract Unveiling an empty street from crowded observations captured by in-car cameras is crucial for autonomous driving. However, removing all temporarily static objects, such as stopped vehicles and standing pedestrians, presents a significant challenge. Unlike object-centric 3D inpain... | [
{
"id": "SmP3hRl1Io",
"initial_rating": 8,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "This paper introduces StreetUnveiler, a method to reconstruct empty street scenes from crowded video captured by moving vehicles. The key challenge is removing te... | {
"rating": "5;6;8;8",
"rating_avg": 6.75,
"confidence": "3;3;4;4",
"confidence_avg": 3.5,
"soundness": "3;3;3;3",
"soundness_avg": 3,
"contribution": "2;3;3;3",
"contribution_avg": 2.75,
"presentation": "3;2;4;3",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.436201"
} | {
"id": "hLKxHTL6KA",
"metareview": "In this paper, the authors have proposed 3D StreetUnveiler to reconstruct an empty street from crowded observations based on semantic 2DGS. It first uses a rendered alpha map to locate unobserved regions, and creates a inpainting mask. Then, a time-reversal inpainting framework ... | {
"decision": "Accept (Poster)"
} |
G6dMvRuhFr | 2411.07223v1 | Grounding Video Models to Actions through Goal Conditioned Exploration | {
"content": "## Abstract\n\nAbstract Large video models, pretrained on massive amounts of Internet video, provide a rich source of physical knowledge about the dynamics and motions of objects and tasks.\nHowever, video models are not grounded in the embodiment of an agent, and do not describe how to actuate the worl... | [
{
"id": "DRnOvuPXMs",
"initial_rating": 8,
"confidence": 3,
"soundness": 3,
"contribution": 4,
"presentation": 4,
"summary": "This paper presents a framework for grounding large pretrained video models to actions within an embodied environment through self-exploration. By generating visu... | {
"rating": "5;6;8",
"rating_avg": 6.333333333333333,
"confidence": "3;3;3",
"confidence_avg": 3,
"soundness": "3;3;3",
"soundness_avg": 3,
"contribution": "4;3;4",
"contribution_avg": 3.6666666666666665,
"presentation": "4;3;4",
"presentation_avg": 3.6666666666666665
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Spotlight",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.436973"
} | {
"id": "KyH76uMrd3",
"metareview": "The paper proposes a novel approach to grounding video models to actions through goal-conditioned exploration, which is a good contribution to the field of embodied AI. The authors have addressed most of the concerns raised by the reviewers in a satisfactory manner. The propose... | {
"decision": "Accept (Spotlight)"
} |
G7u4ue6ncT | 2405.14660v1 | Implicit In-context Learning | {
"content": "## Abstract\n\nAbstract In-context Learning (ICL) empowers large language models (LLMs) to adapt to unseen tasks during inference by prefixing a few demonstration examples prior to test queries. Despite its versatility, ICL incurs substantial computational and memory overheads compared to zero-shot lear... | [
{
"id": "4MfGItNk1h",
"initial_rating": 6,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "The paper I2CL proposes a method to achieve few-shot performance at zero-shot inference costs by generating a condensed \"context vector\" from demonstration exam... | {
"rating": "5;5;6;8",
"rating_avg": 6,
"confidence": "3;3;3;4",
"confidence_avg": 3.25,
"soundness": "4;3;3;3",
"soundness_avg": 3.25,
"contribution": "2;2;3;3",
"contribution_avg": 2.5,
"presentation": "3;4;3;3",
"presentation_avg": 3.25
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.437771"
} | {
"id": "WB8TCjmDKO",
"metareview": "This work proposes a new implicit in-context learning that extracts a context vector from demonstration examples and injects the context vector into the model’s residual stream to reduce the inference cost of ICL to that of zero-shot learning All reviewers consistently recommen... | {
"decision": "Accept (Poster)"
} |
G82uQztzxl | 2410.20098v1 | Self-Normalized Resets for Plasticity in Continual Learning | {
"content": "## Abstract\n\nAbstract Plasticity Loss is an increasingly important phenomenon that refers to the empirical observation that as a neural network is continually trained on a sequence of changing tasks, its ability to adapt to a new task diminishes over time. We introduce Self-Normalized Resets (SNR), a ... | [
{
"id": "iWc2lk7hgo",
"initial_rating": 5,
"confidence": 3,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "The paper proposes a particular heuristic (SNR) to identify neurons that need to be reset in order to recover plasticity. They show that this particular choice of... | {
"rating": "3;3;5;5;8",
"rating_avg": 4.8,
"confidence": "3;4;4;3;3",
"confidence_avg": 3.4,
"soundness": "2;2;2;3;4",
"soundness_avg": 2.6,
"contribution": "2;2;2;2;4",
"contribution_avg": 2.4,
"presentation": "3;1;2;3;3",
"presentation_avg": 2.4
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.438673"
} | {
"id": "TOA1JzHnIy",
"metareview": "The paper introduces Self-Normalised Resets (SNR), a novel method to mitigate plasticity loss in continual learning by resetting neurons. The method is simple and requires minimal parameter tuning, yet it outperforms competing methods on standard benchmarks. The authors provide ... | {
"decision": "Accept (Poster)"
} |
G84F1h2IiD | 2312.16903v3 | Spike No More: Stabilizing the Pre-training of Large Language Models | {
"content": "## Abstract\n\nAbstract Loss spikes often occur during pre-training of large language models.\nThe spikes degrade the performance of large language models and sometimes ruin the pre-training.\nSince the pre-training needs a vast computational budget, we should avoid such spikes.\nBased on the assumption... | [
{
"id": "FNlKyP0Y4O",
"initial_rating": 3,
"confidence": 3,
"soundness": 2,
"contribution": 2,
"presentation": 2,
"summary": "This paper presents a strategy to avoid spikes in loss during the training of LMs by keeping the gradient norm small. To manage the upper limit of gradient norms ... | {
"rating": "3;3;5",
"rating_avg": 3.6666666666666665,
"confidence": "3;3;4",
"confidence_avg": 3.3333333333333335,
"soundness": "3;2;2",
"soundness_avg": 2.3333333333333335,
"contribution": "2;2;2",
"contribution_avg": 2,
"presentation": "3;2;2",
"presentation_avg": 2.3333333333333335
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:01.439678"
} | {
"id": "jUsfiJJxWA",
"metareview": "The paper analyses the important question of divergence in optimizing the parameters of large neural language models. The authors investigate the high gradient norms as the cause for loss spikes, and discuss two mitigation strategies in terms of having small sub-layers and large... | {
"decision": "Reject"
} |
G8U2nGP3Vi | 2406.02502v1 | Singular Subspace Perturbation Bounds via Rectangular Random Matrix Diffusions | {
"content": "## Abstract\n\nAbstract Given a matrix A ∈ ℝ m × d 𝐴 superscript ℝ 𝑚 𝑑 A\\in\\mathbb{R}^{m\\times d} italic_A ∈ blackboard_R start_POSTSUPERSCRIPT italic_m × italic_d end_POSTSUPERSCRIPT with singular values σ 1 ≥ ⋯ ≥ σ d subscript 𝜎 1 ⋯ subscript 𝜎 𝑑 \\sigma_{1}\\geq\\cdots\\geq\\sigma_{d} italic... | [
{
"id": "jk47vmEDXK",
"initial_rating": 5,
"confidence": 3,
"soundness": 3,
"contribution": 2,
"presentation": 2,
"summary": "The paper provides a new upper-bound for the right singular vectors of a perturbed matrix by Gaussian noise. The authors show that under a set of extra assumption... | {
"rating": "3;3;5;6;6",
"rating_avg": 4.6,
"confidence": "5;3;3;3;2",
"confidence_avg": 3.2,
"soundness": "3;3;3;3;3",
"soundness_avg": 3,
"contribution": "1;3;2;3;3",
"contribution_avg": 2.4,
"presentation": "3;3;2;3;4",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.440501"
} | {
"id": "pMx1aBnQ29",
"metareview": "This paper provides new matrix perturbation bounds for the top singular vectors of a matrix A under Gaussian perturbations. The authors improve prior results in the setting when the matrix is much taller than it is wide (a common setting in data applications). Most of the review... | {
"decision": "Accept (Poster)"
} |
G9xhvGPtte | 2410.17434v1 | LongVU: Spatiotemporal Adaptive Compression for Long Video-Language Understanding | {
"content": "## Abstract\n\nAbstract Multimodal Large Language Models (MLLMs) have shown promising progress in understanding and analyzing video content. However, processing long videos remains a significant challenge constrained by LLM’s context size. To address this limitation, we propose LongVU , a spatiotemporal... | [
{
"id": "h5ibF3Wwy7",
"initial_rating": 6,
"confidence": 4,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "The paper presents LongVU, a spatiotemporal adaptive compression mechanism designed to handle long video sequences within the limited context length constraints o... | {
"rating": "5;5;6;6;6",
"rating_avg": 5.6,
"confidence": "5;4;4;4;4",
"confidence_avg": 4.2,
"soundness": "3;3;3;3;3",
"soundness_avg": 3,
"contribution": "2;2;3;3;2",
"contribution_avg": 2.4,
"presentation": "3;3;4;3;3",
"presentation_avg": 3.2
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:01.441334"
} | {
"id": "KRKZngSGW5",
"metareview": "This paper proposes LongVU, a spatiotemporal adaptive compression mechanism for long video-language understanding. It aims to address the challenge of limited context length in current multimodal large language models (MLLMs) by reducing video tokens while preserving visual deta... | {
"decision": "Reject"
} |
GARbxyCV13 | 2411.04983v1 | DINO-WM: World Models on Pre-trained Visual Features enable Zero-shot Planning | {
"content": "## Abstract\n\nAbstract The ability to predict future outcomes given control actions is fundamental for physical reasoning. However, such predictive models, often called world models, have proven challenging to learn and are typically developed for task-specific solutions with online policy learning. We... | [
{
"id": "ad0QT9Ln28",
"initial_rating": 3,
"confidence": 4,
"soundness": 3,
"contribution": 2,
"presentation": 2,
"summary": "This paper introduces a visual dynamics learning approach called DINO-WM, leveraging pre-trained DINOv2 latent features for predictive modeling. DINO-WM incorpora... | {
"rating": "3;5;6;6",
"rating_avg": 5,
"confidence": "4;5;4;4",
"confidence_avg": 4.25,
"soundness": "3;3;3;3",
"soundness_avg": 3,
"contribution": "2;2;3;2",
"contribution_avg": 2.25,
"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": "2026-01-14T22:16:01.442006"
} | {
"id": "mLsKIJajoV",
"metareview": "The paper presents DINO-WM, a method for learning world models that leverage pre-trained DINOv2 visual features to enable zero-shot planning. The core idea is to predict future states in the latent space of DINOv2's patch features, trained purely from offline trajectories withou... | {
"decision": "Reject"
} |
GBWqZNoeIk | 2410.08125v1 | Generalizing Stochastic Smoothing for Differentiation and Gradient Estimation | {
"content": "## Abstract\n\nAbstract We deal with the problem of gradient estimation for stochastic differentiable relaxations of algorithms, operators, simulators, and other non-differentiable functions.\nStochastic smoothing conventionally perturbs the input of a non-differentiable function with a differentiable d... | [
{
"id": "OrQaifyGFP",
"initial_rating": 3,
"confidence": 5,
"soundness": 4,
"contribution": 1,
"presentation": 4,
"summary": "This work presents methods for stochastic smoothing of non-differentiable functions, motivated by applications where there is the need to optimize non-differentia... | {
"rating": "3;3;6;8",
"rating_avg": 5,
"confidence": "3;5;3;3",
"confidence_avg": 3.5,
"soundness": "2;4;3;3",
"soundness_avg": 3,
"contribution": "1;1;3;3",
"contribution_avg": 2,
"presentation": "2;4;4;4",
"presentation_avg": 3.5
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:01.442867"
} | {
"id": "OTUSV89YR8",
"metareview": "While the paper aims to address interesting problems, several reviewers have found its theoretical contribution, in light of the existing literature, to be rather minimal, with its impact on machine learning deemed questionable. It has been noted that the theoretical derivations... | {
"decision": "Reject"
} |
GDd5H92egZ | 2407.12877v2 | ReFeR: Improving Evaluation and Reasoning through Hierarchy of Models | {
"content": "## Abstract\n\nAbstract Assessing the quality of outputs generated by generative models, such as large language models and vision language models, presents notable challenges. Traditional methods for evaluation typically rely on either human assessments, which are resource-intensive, or automatic metric... | [
{
"id": "DOh7T9VUf3",
"initial_rating": 5,
"confidence": 4,
"soundness": 2,
"contribution": 2,
"presentation": 3,
"summary": "The paper proposes a hierarchical multi-LLM framework using a peer review structure for text evaluation and reasoning. The framework employs smaller models as pee... | {
"rating": "3;3;5;5;6",
"rating_avg": 4.4,
"confidence": "4;4;3;4;3",
"confidence_avg": 3.6,
"soundness": "2;2;2;2;3",
"soundness_avg": 2.2,
"contribution": "3;1;2;2;2",
"contribution_avg": 2,
"presentation": "3;3;2;3;3",
"presentation_avg": 2.8
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:01.445921"
} | {
"id": "PKwtMcGOCr",
"metareview": "This paper received ratings of 6, 6, 5, 5, 5, and was recommended for rejection by majority of reviewers.\n\nThe paper introduces ReFeR, a hierarchical multi-agent framework inspired by the academic peer review process. It utilizes multiple LLMs or VLMs as peer evaluators and a ... | {
"decision": "Reject"
} |
GFgn2LprFR | 2411.01894v1 | Efficient Active Imitation Learning with Random Network Distillation | {
"content": "## Abstract\n\nAbstract Developing agents for complex and underspecified tasks, where no clear objective exists, remains challenging but offers many opportunities. This is especially true in video games, where simulated players (bots) need to play realistically, and there is no clear reward to evaluate ... | [
{
"id": "FwIg1R0FkO",
"initial_rating": 6,
"confidence": 4,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "This work presents a method to more effectively integrate human feedback with imitation learning and specifically the IL method dagger. The idea is to only reques... | {
"rating": "5;5;5;6",
"rating_avg": 5.25,
"confidence": "3;3;4;4",
"confidence_avg": 3.5,
"soundness": "2;2;3;3",
"soundness_avg": 2.5,
"contribution": "3;2;2;2",
"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": "2026-01-14T22:16:01.447328"
} | {
"id": "mXtNIM2wYX",
"metareview": "(a) Summary: This paper presents Random Network Distillation DAgger (RND-DAgger), an active imitation learning method that leverages RND to define out-of-distribution states, enabling selective expert intervention and minimizing the frequency of transitions between human experts... | {
"decision": "Accept (Poster)"
} |
GFua0WEYGF | 2410.19931v2 | Provable optimal transport with transformers: The essence of depth and prompt engineering | {
"content": "## Abstract\n\nAbstract Can we establish provable performance guarantees for transformers? Establishing such theoretical guarantees is a milestone in developing trustworthy generative AI. In this paper, we take a step toward addressing this question by focusing on optimal transport, a fundamental proble... | [
{
"id": "qIPyGjxzhS",
"initial_rating": 5,
"confidence": 4,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "This article constructs a prompt that enables transformers to perform gradient descent on dual optimal transport. Additionally, it provides an explicit approximat... | {
"rating": "3;5;5;5",
"rating_avg": 4.5,
"confidence": "4;3;2;4",
"confidence_avg": 3.25,
"soundness": "4;3;3;3",
"soundness_avg": 3.25,
"contribution": "2;2;3;2",
"contribution_avg": 2.25,
"presentation": "1;2;2;3",
"presentation_avg": 2
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:01.448015"
} | {
"id": "7MEoZ6Qwdm",
"metareview": "In this paper, the authors show that there exists a particular prompt with a specific configuration of parameters for transformers that implement Sinkhorn's algorithm and thus approximatively solve regularized optimal transport problems. \n\nHaving looked at the paper and the re... | {
"decision": "Reject"
} |
GG80jy9KI5 | 2408.05690v1 | Strong denoising of financial time-series | {
"content": "## Abstract\n\nAbstract In this paper we introduce a method for significantly improving the signal to noise ratio in financial data. The approach relies on combining a target variable with different context variables and use auto-encoders (AEs) to learn reconstructions of the combined inputs. The object... | [
{
"id": "LD7pGw4GvN",
"initial_rating": 5,
"confidence": 2,
"soundness": 2,
"contribution": 2,
"presentation": 1,
"summary": "This paper proposes a method for learning denoised representations of financial time series. They use distinct autoencoders (AEs) that mutually regularize each ot... | {
"rating": "3;3;3;5",
"rating_avg": 3.5,
"confidence": "3;3;2;2",
"confidence_avg": 2.5,
"soundness": "1;2;1;2",
"soundness_avg": 1.5,
"contribution": "2;1;1;2",
"contribution_avg": 1.5,
"presentation": "1;1;2;1",
"presentation_avg": 1.25
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:01.448653"
} | {
"id": "J6n2Zkm75j",
"metareview": "The paper introduces a method for learning denoised representations of financial time series. It leverages dual autoencoders trained on a reconstruction task, where they mutually refine their reconstructions through a collaborative process. The autoencoders are constrained to ag... | {
"decision": "Reject"
} |
GGAG3wFEKv | 2402.10028v1 | Diffusion Models Meet Contextual Bandits | {
"content": "## Abstract\n\nAbstract Efficient exploration is a key challenge in contextual bandits due to the large size of their action space, where uninformed exploration can result in computational and statistical inefficiencies. Fortunately, the rewards of actions are often correlated and this can be leveraged ... | [
{
"id": "lqK8IZsnix",
"initial_rating": 6,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "This paper introduces Diffusion Thompson Sampling (dTS), a novel algorithm leveraging pre-trained diffusion model priors to optimize exploration in contextual ban... | {
"rating": "3;5;5;6;6",
"rating_avg": 5,
"confidence": "2;3;4;5;4",
"confidence_avg": 3.6,
"soundness": "2;3;2;3;3",
"soundness_avg": 2.6,
"contribution": "2;2;2;2;3",
"contribution_avg": 2.2,
"presentation": "2;3;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": "2026-01-14T22:16:01.449493"
} | {
"id": "VJlWnFEHmY",
"metareview": "This paper introduces Diffusion Thompson Sampling (dTS), an algorithm that addresses the exploration-exploitation trade-off in contextual bandits with large action spaces. By integrating pre-trained diffusion models, dTS identifies correlations between actions, facilitating more... | {
"decision": "Reject"
} |
GJsuYHhAga | 2410.08261v1 | Meissonic: Revitalizing Masked Generative Transformers for Efficient High-Resolution Text-to-Image Synthesis | {
"content": "## Abstract\n\nAbstract Diffusion models, such as Stable Diffusion, have made significant strides in visual generation, yet their paradigm remains fundamentally different from autoregressive language models, complicating the development of unified language-vision models. Recent efforts like LlamaGen hav... | [
{
"id": "1IoJd9L5LP",
"initial_rating": 6,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 4,
"summary": "- The paper presents Meissonic, a high-resolution, non-autoregressive masked image modeling (MIM) approach for text to image generation. \n- It introduces archite... | {
"rating": "3;6;6",
"rating_avg": 5,
"confidence": "5;4;4",
"confidence_avg": 4.333333333333333,
"soundness": "2;3;3",
"soundness_avg": 2.6666666666666665,
"contribution": "2;2;3",
"contribution_avg": 2.3333333333333335,
"presentation": "2;4;4",
"presentation_avg": 3.3333333333333335
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.450460"
} | {
"id": "ksA5jFEuk4",
"metareview": "The work introduced Messonic that improved MIM with enhanced transformer architecture, advanced positional encoding and adding masking rate as a condition. The effectiveness of proposed approach is demonstrated well and efficiency brought by the improvement is obvious. It consis... | {
"decision": "Accept (Poster)"
} |
GKAQ92ua3A | 2405.03233v2 | ADMM for Nonconvex Optimization under Minimal Continuity Assumption | {
"content": "## Abstract\n\nAbstract This paper introduces a novel approach to solving multi-block nonconvex composite optimization problems through a proximal linearized Alternating Direction Method of Multipliers (ADMM). This method incorporates an Increasing Penalization and Decreasing Smoothing (IPDS) strategy. ... | [
{
"id": "t4YmnnLA5v",
"initial_rating": 5,
"confidence": 3,
"soundness": 3,
"contribution": 2,
"presentation": 2,
"summary": "This paper investigates the ADMM for multi-block composite optimization problems. An adaptive penalization technique is employed to enhance convergence. The globa... | {
"rating": "5;5;5",
"rating_avg": 5,
"confidence": "2;3;3",
"confidence_avg": 2.6666666666666665,
"soundness": "3;3;3",
"soundness_avg": 3,
"contribution": "2;2;2",
"contribution_avg": 2,
"presentation": "3;2;2",
"presentation_avg": 2.3333333333333335
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.451545"
} | {
"id": "rLlQakBbZ9",
"metareview": "This paper designs ADMM for solving multi-block nonconvex composite optimization problems. The proposed approach requires less stringent conditions comparing with existing work. An improved oracle complexity is also established for the proposed method. Numerical results demonstr... | {
"decision": "Accept (Poster)"
} |
GMwRl2e9Y1 | 2410.06424v1 | Restructuring Vector Quantization with the Rotation Trick | {
"content": "## Abstract\n\nAbstract Vector Quantized Variational AutoEncoders (VQ-VAEs) are designed to compress a continuous input to a discrete latent space and reconstruct it with minimal distortion.\nThey operate by maintaining a set of vectors—often referred to as the codebook—and quantizing each encoder outpu... | [
{
"id": "DpDr22igpj",
"initial_rating": 8,
"confidence": 4,
"soundness": 4,
"contribution": 3,
"presentation": 4,
"summary": "This paper proposes an improvement to the STE method for training VQ-VAE. While STE copies the gradient from the quantizer output to input, this paper proposes th... | {
"rating": "3;5;6;8",
"rating_avg": 5.5,
"confidence": "4;4;3;4",
"confidence_avg": 3.75,
"soundness": "4;3;3;3",
"soundness_avg": 3.25,
"contribution": "2;3;3;3",
"contribution_avg": 2.75,
"presentation": "3;2;3;4",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Oral",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.452800"
} | {
"id": "CFMVNzuBmQ",
"metareview": "All reviewers appreciate the paper’s contribution and strongly recommend accepting the paper. The AC concurs, after reading the paper quickly.",
"additional_comments": "NA"
} | {
"decision": "Accept (Oral)"
} |
GNOMC90vbl | 2406.19622v1 | Data-Driven Lipschitz Continuity: A Cost-Effective Approach to Improve Adversarial Robustness | {
"content": "## Abstract\n\nAbstract The security and robustness of deep neural networks (DNNs) have become increasingly concerning. This paper aims to provide both a theoretical foundation and a practical solution to ensure the reliability of DNNs. We explore the concept of Lipschitz continuity to certify the robus... | [
{
"id": "Q9LOOAbTCB",
"initial_rating": 5,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "This work starts from Lipschitz continuity and enhances adversarial robustness by minimizing the empirical Lipschitz constant. Specifically, the authors propose a... | {
"rating": "3;3;5;5;5",
"rating_avg": 4.2,
"confidence": "3;4;4;4;4",
"confidence_avg": 3.8,
"soundness": "2;1;2;3;2",
"soundness_avg": 2,
"contribution": "2;1;2;3;2",
"contribution_avg": 2,
"presentation": "2;2;3;3;2",
"presentation_avg": 2.4
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:01.453603"
} | {
"id": "mcUrxQtQsx",
"metareview": "The paper proposes a method to enhance adversarial robustness in deep neural networks by reducing the empirical Lipschitz constant through a novel forged function applied during inference. While the work aims to address a critical issue in AI safety and robustness, it falls shor... | {
"decision": "Reject"
} |
GPDcvoFGOL | 2406.04341v2 | Interpreting the Second-Order Effects of Neurons in CLIP | {
"content": "## Abstract\n\nAbstract We interpret the function of individual neurons in CLIP by automatically describing them using text. Analyzing the direct effects (i.e. the flow from a neuron through the residual stream to the output) or the indirect effects (overall contribution) fails to capture the neurons’ f... | [
{
"id": "O6demcC0Sc",
"initial_rating": 5,
"confidence": 4,
"soundness": 3,
"contribution": 2,
"presentation": 2,
"summary": "This work presents a novel approach for examining potential second-order effects of neurons of CLIP representations and how these can be used in the context of ze... | {
"rating": "5;6;8;8",
"rating_avg": 6.75,
"confidence": "4;3;3;4",
"confidence_avg": 3.5,
"soundness": "3;2;3;3",
"soundness_avg": 2.75,
"contribution": "2;2;3;4",
"contribution_avg": 2.75,
"presentation": "2;2;3;3",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.454396"
} | {
"id": "wyMgoHoVOW",
"metareview": "The paper introduces a novel method for interpreting neurons in CLIP (Contrastive Language-Image Pretraining) through second-order effects, which measure a neuron’s influence on the model’s output via later attention layers. This approach provides a deeper understanding of indiv... | {
"decision": "Accept (Poster)"
} |
GQ1Tc3vHbt | 2410.10800v1 | Optimizing $(L_0, L_1)$-Smooth Functions by Gradient Methods | {
"content": "## 1 Introduction\n\nIn this paper, we focus on the deterministic unconstrained optimization problem\n\n$$ $f^{*}\\coloneq\\min_{x\\in\\mathbb{R}^{d}}f(x),$ (1) $$\n\nwhere $f\\colon\\mathbb{R}^{d}\\to\\mathbb{R}$ is a smooth function.\nWith the rise of deep learning, ensuring efficient convergence has ... | [
{
"id": "arkYVX48mO",
"initial_rating": 6,
"confidence": 3,
"soundness": 4,
"contribution": 3,
"presentation": 2,
"summary": "This paper provides a general understanding of the $(L_0, L_1)$-smooth function class, which has recently gained increasing interest due to its empirically observ... | {
"rating": "5;5;6;6",
"rating_avg": 5.5,
"confidence": "5;4;3;3",
"confidence_avg": 3.75,
"soundness": "3;3;3;4",
"soundness_avg": 3.25,
"contribution": "2;2;3;3",
"contribution_avg": 2.5,
"presentation": "3;3;3;2",
"presentation_avg": 2.75
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.455094"
} | {
"id": "1LbvBFq2Zv",
"metareview": "The paper studies gradient methods for optimization problems with \\((L_0, L_1)\\)-smooth objectives, a generalization of Lipschitz-smooth functions. It offers new insights into this class and introduces a unified analysis framework for related methods. The authors establish sta... | {
"decision": "Accept (Poster)"
} |
GRMfXcAAFh | 2410.03943v1 | Oscillatory State-Space Models | {
"content": "## Abstract\n\nAbstract We propose Linear Oscillatory State-Space models (LinOSS) for efficiently learning on long sequences. Inspired by cortical dynamics of biological neural networks, we base our proposed LinOSS model on a system of forced harmonic oscillators. A stable discretization, integrated ove... | [
{
"id": "0kU14BMTLi",
"initial_rating": 5,
"confidence": 5,
"soundness": 3,
"contribution": 3,
"presentation": 4,
"summary": "This paper introduces a new continuous time recurrent network in the family of state space models. The architecture is proposed as an ODE that is discretized in t... | {
"rating": "5;6;6;8",
"rating_avg": 6.25,
"confidence": "5;3;4;2",
"confidence_avg": 3.5,
"soundness": "3;3;3;4",
"soundness_avg": 3.25,
"contribution": "3;2;3;4",
"contribution_avg": 3,
"presentation": "4;3;3;3",
"presentation_avg": 3.25
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Oral",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.455604"
} | {
"id": "awZtadAfy6",
"metareview": "It is a pleasure when the reviewers all converge on a consistent recommendation. I see no reason to disagree with the universal praise for the paper, and strongly recommend acceptance, and believe the paper should be highlighted with a spotlight presentation.",
"additional_com... | {
"decision": "Accept (Oral)"
} |
GSUNPIw7Ad | 2407.19651v1 | Bridging Compressed Image Latents and Multimodal Large Language Models | {
"content": "## Abstract\n\nAbstract This paper presents the first-ever study of adapting compressed image latents to suit the needs of downstream vision tasks that adopt Multimodal Large Language Models (MLLMs). MLLMs have extended the success of large language models to modalities (e.g. images) beyond text, but th... | [
{
"id": "CihjUQVDhm",
"initial_rating": 6,
"confidence": 4,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "This paper suggests using compressed image latents for more efficient multimodal large language models (MLLMs). Specifically, a lightweight transform-neck and a s... | {
"rating": "3;5;6;6",
"rating_avg": 5,
"confidence": "2;3;3;3",
"confidence_avg": 2.75,
"soundness": "2;2;3;3",
"soundness_avg": 2.5,
"contribution": "2;2;3;2",
"contribution_avg": 2.25,
"presentation": "2;3;3;3",
"presentation_avg": 2.75
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.456260"
} | {
"id": "E5Z24sczVq",
"metareview": "This paper studies using neural image compression to reduce the transmission rate between devices and servers for Multimodal Large Language Models. The proposed method involves a lightweight transform-neck, which is trained with surrogate loss to adapt image latents for MLLMs. T... | {
"decision": "Accept (Poster)"
} |
GTe9PDhm8v | 2410.19103v1 | TesseraQ: Ultra Low-Bit LLM Post-Training Quantization with Block Reconstruction | {
"content": "## Abstract\n\nAbstract Large language models (LLMs) have revolutionized natural language processing, albeit at the cost of immense memory and computation requirements. Post-training quantization (PTQ) is becoming the de facto method to reduce the memory footprint and improve the inference throughput of... | [
{
"id": "DUX824QwQ7",
"initial_rating": 5,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 2,
"summary": "This paper presents a PTQ method，aTesseraQ, for LLM. The author propose optimize weight rounding through a progressive\napproach based on block reconstruction. B... | {
"rating": "3;5;5;6;6",
"rating_avg": 5,
"confidence": "5;4;3;3;5",
"confidence_avg": 4,
"soundness": "2;2;3;3;3",
"soundness_avg": 2.6,
"contribution": "2;2;3;3;2",
"contribution_avg": 2.4,
"presentation": "3;2;2;4;3",
"presentation_avg": 2.8
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:01.457036"
} | {
"id": "G1wde3xvU9",
"metareview": "The paper introduces TesseraQ, a novel post-training quantization (PTQ) technique for large language models (LLMs), aiming to improve performance by optimizing weight rounding through a block reconstruction approach. The authors also propose a progressive adaptive rounding strat... | {
"decision": "Reject"
} |
GULx8rzzjC | 2410.08432v1 | Mycroft: Towards Effective and Efficient External Data Augmentation | {
"content": "## Abstract\n\nAbstract Machine learning (ML) models often require large amounts of data to perform\nwell. When the data available to the model trainer is insufficient to obtain\ngood performance on their desired task, they may need to acquire more data\nfrom external sources. Often, useful data is held... | [
{
"id": "ZD872g5n5B",
"initial_rating": 5,
"confidence": 3,
"soundness": 4,
"contribution": 3,
"presentation": 3,
"summary": "The paper introduces \"Mycroft,\" a method designed to enable model trainers to assess the utility of external data sources with limited data sharing, addressing ... | {
"rating": "3;3;3;5",
"rating_avg": 3.5,
"confidence": "3;3;3;3",
"confidence_avg": 3,
"soundness": "2;3;2;4",
"soundness_avg": 2.75,
"contribution": "1;2;2;3",
"contribution_avg": 2,
"presentation": "2;3;2;3",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:01.457800"
} | {
"id": "CU0NrYSOJf",
"metareview": "The reviewers were not excited about the paper. Additionally the authors did not submit rebuttals to clarify the concerns of the reviewers. There are a bunch of problem formulation and algorithm description concerns that require addressing. We would recommend the authors to addr... | {
"decision": "Reject"
} |
GVNYi74t5L | 2405.15638v1 | M4U: Evaluating Multilingual Understanding and Reasoning for Large Multimodal Models | {
"content": "## Abstract\n\nAbstract Multilingual multimodal reasoning is a core component in achieving human-level intelligence. However, most existing benchmarks for multilingual multimodal reasoning struggle to differentiate between models of varying performance; even language models without visual capabilities c... | [
{
"id": "vzyrOakb0N",
"initial_rating": 6,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "This paper curates an evaluation benchmark (called M4U) for multilingual multimodal models. It consists of 8,931 samples covering 64 disciplines across 16 subfiel... | {
"rating": "3;3;5;6",
"rating_avg": 4.25,
"confidence": "4;4;4;4",
"confidence_avg": 4,
"soundness": "2;2;2;3",
"soundness_avg": 2.25,
"contribution": "2;2;3;3",
"contribution_avg": 2.5,
"presentation": "1;2;3;3",
"presentation_avg": 2.25
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:01.458491"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
GYk0thSY1M | 2406.06110v1 | Recurrent Context Compression: Efficiently Expanding the Context Window of LLM | {
"content": "## Abstract\n\nAbstract To extend the context length of Transformer-based large language models (LLMs) and improve comprehension capabilities, we often face limitations due to computational resources and bounded memory storage capacity. This work introduces a method called Recurrent Context Compression ... | [
{
"id": "NBwjE2lSgi",
"initial_rating": 5,
"confidence": 4,
"soundness": 2,
"contribution": 2,
"presentation": 3,
"summary": "The paper introduces Recurrent Context Compression (RCC), a technique for long-context language modeling. The input is segmented into smaller chunks, the encoder ... | {
"rating": "3;3;5;5",
"rating_avg": 4,
"confidence": "3;4;4;4",
"confidence_avg": 3.75,
"soundness": "2;3;2;2",
"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": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:01.459130"
} | {
"id": "dkHosY8cRU",
"metareview": "Based on the reviewers' feedback, I recommend not to accept this paper at this time. While the Recurrent Context Compression (RCC) approach presents an interesting direction for extending LLM context length, several critical issues remain: limited novelty compared to prior work ... | {
"decision": "Reject"
} |
GbXn0Dgf7f | 2302.00098v1 | Does Deep Active Learning Work in the Wild? | {
"content": "## Abstract\n\nAbstract Deep learning (DL) is revolutionizing the scientific computing community. To reduce the data gap caused by usually expensive simulations or experimentation, active learning has been identified as a promising solution for the scientific computing community. However, the deep activ... | [
{
"id": "Z6Ddifto7I",
"initial_rating": 3,
"confidence": 5,
"soundness": 2,
"contribution": 2,
"presentation": 2,
"summary": "This is a benchmark paper, the author evaluates eleven state-of-the-art DAL methods across eight datasets in the wild.",
"strengths": "1. It tests eleven diff... | {
"rating": "3;3;3;3;5",
"rating_avg": 3.4,
"confidence": "3;3;3;5;4",
"confidence_avg": 3.6,
"soundness": "2;2;1;2;2",
"soundness_avg": 1.8,
"contribution": "2;2;1;2;2",
"contribution_avg": 1.8,
"presentation": "3;2;2;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": "2026-01-14T22:16:01.459920"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
GbgCRJedQ7 | 2405.15525v2 | Sparse Matrix in Large Language Model Fine-tuning | {
"content": "## Abstract\n\nAbstract LoRA and its variants have become popular parameter-efficient fine-tuning (PEFT) methods due to their ability to avoid excessive computational costs.\nHowever, an accuracy gap often exists between PEFT methods and full fine-tuning (FT), and this gap has yet to be systematically s... | [
{
"id": "rwfb8grzdw",
"initial_rating": 6,
"confidence": 2,
"soundness": 3,
"contribution": 3,
"presentation": 2,
"summary": "This paper presents a novel PEFT method, Sparse Matrix Tuning. SMT uses a gradient based criteria to identify the most relevant submatrices of the model to update... | {
"rating": "5;5;5;6;8",
"rating_avg": 5.8,
"confidence": "4;4;3;2;4",
"confidence_avg": 3.4,
"soundness": "3;2;3;3;3",
"soundness_avg": 2.8,
"contribution": "2;2;2;3;3",
"contribution_avg": 2.4,
"presentation": "3;2;2;2;4",
"presentation_avg": 2.6
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.460750"
} | {
"id": "ayL17xqJNP",
"metareview": "This paper introduces Sparse Matrix Tuning (SMT), a parameter-efficient fine-tuning method for large language models. By selectively updating the most significant submatrices based on gradient analysis, SMT reduces computational and memory costs while maintaining performance lev... | {
"decision": "Accept (Poster)"
} |
Gc2qkiYUkh | 2410.08194v1 | Features are fate: a theory of transfer learning in high-dimensional regression | {
"content": "## Abstract\n\nAbstract With the emergence of large-scale pre-trained neural networks, methods to adapt such “foundation” models to data-limited downstream tasks have become a necessity.\nFine-tuning, preference optimization, and transfer learning have all been successfully employed for these purposes w... | [
{
"id": "lxGJIlnwrs",
"initial_rating": 6,
"confidence": 2,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "This paper addresses the gap in understanding how to adapt large pre-trained models to data-limited tasks by examining the theory behind task similarity in transf... | {
"rating": "3;5;6;6;6",
"rating_avg": 5.2,
"confidence": "3;4;2;3;2",
"confidence_avg": 2.8,
"soundness": "2;2;2;3;3",
"soundness_avg": 2.4,
"contribution": "2;2;3;3;3",
"contribution_avg": 2.6,
"presentation": "2;4;2;4;3",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:01.461523"
} | {
"id": "dK46fP8Enu",
"metareview": "This paper studies under a new feature-centric viewpoint the similarity of source and target tasks in transfer learning. The implications of such an analysis should be important because they can help understand the intricacies of knowledge transfer from a foundation model to a n... | {
"decision": "Reject"
} |
GcJE0HPy4X | 2408.11338v1 | Automatic Dataset Construction (ADC): Sample Collection, Data Curation, and Beyond | {
"content": "## Abstract\n\nAbstract Large-scale data collection is essential for developing personalized training data, mitigating the shortage of training data, and fine-tuning specialized models. However, creating high-quality datasets quickly and accurately remains a challenge due to annotation errors, the subst... | [
{
"id": "ksIiELmbIj",
"initial_rating": 8,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "The paper tackles the problem of automatic dataset creation. It proposes the ADC pipeline for the same which requires minimal human overhead. The proposal is to f... | {
"rating": "3;5;6;8",
"rating_avg": 5.5,
"confidence": "4;4;3;3",
"confidence_avg": 3.5,
"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": "2026-01-14T22:16:01.462344"
} | {
"id": "7m6Lkmstdl",
"metareview": "This paper was reviewed by four experts in the field and received 8, 5, 5, 6 as the final ratings. The reviewers agreed that the paper tackles an important problem of dataset creation, it is well-written and easy to follow, and the hyper-parameter details are clearly specified. ... | {
"decision": "Reject"
} |
GeTBk67mK6 | 2410.04509v2 | ErrorRadar: Benchmarking Complex Mathematical Reasoning of Multimodal Large Language Models Via Error Detection | {
"content": "## Abstract\n\nAbstract As the field of Multimodal Large Language Models (MLLMs) continues to evolve, their potential to revolutionize artificial intelligence is particularly promising, especially in addressing mathematical reasoning tasks. Current mathematical benchmarks predominantly focus on evaluati... | [
{
"id": "ik8OPHmqmN",
"initial_rating": 5,
"confidence": 4,
"soundness": 2,
"contribution": 2,
"presentation": 2,
"summary": "This paper presents ERRORRADAR, an innovative benchmark aimed at evaluating the capabilities of Multimodal Large Language Models (MLLMs) in mathematical error det... | {
"rating": "5;5;5;6",
"rating_avg": 5.25,
"confidence": "3;5;4;3",
"confidence_avg": 3.75,
"soundness": "2;2;2;4",
"soundness_avg": 2.5,
"contribution": "3;3;2;3",
"contribution_avg": 2.75,
"presentation": "2;3;2;3",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:01.463084"
} | {
"id": "81Kt2higy5",
"metareview": "This was a borderline paper that I believe was submitted too soon. The reviewers pointed out extensive issues with the paper and while the authors did significant work in addressing those issues during the rebuttal period, the main result was that one reviewer moved from margin... | {
"decision": "Reject"
} |
GfXMTAJaxZ | 2409.06594v1 | How to Verify Any (Reasonable) Distribution Property: Computationally Sound Argument Systems for Distributions | {
"content": "## Abstract\n\nAbstract As statistical analyses become more central to science, industry and society, there is a growing need to ensure correctness of their results. Approximate correctness can be verified by replicating the entire analysis, but can we verify without replication? Building on a recent li... | [
{
"id": "mBFaK8PZgd",
"initial_rating": 6,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 2,
"summary": "This paper falls in the area of distribution testing. In the standard setting of distribution testing, we have given sample access to an unknown distribution $D$ ... | {
"rating": "5;6;6;6;8;8;8",
"rating_avg": 6.714285714285714,
"confidence": "3;4;4;3;3;3;3",
"confidence_avg": 3.2857142857142856,
"soundness": "3;3;3;3;4;3;4",
"soundness_avg": 3.2857142857142856,
"contribution": "3;2;2;2;3;3;4",
"contribution_avg": 2.7142857142857144,
"presentation": "2;3;3;3;4;2;4"... | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.464075"
} | {
"id": "GchbezlFGp",
"metareview": "This paper considers the problem of verifying whether an unknown discrete distribution satisfies certain properties with a prover. The authors show that it is possible to verify a wide class of properties using a sublinear number of samples and running time. Without a prover, ev... | {
"decision": "Accept (Poster)"
} |
GhM63V7z6v | 2409.19635v1 | Temporal Source Recovery for Time-Series Source-Free Unsupervised Domain Adaptation | {
"content": "## Abstract\n\nAbstract Source-Free Unsupervised Domain Adaptation (SFUDA) has gained popularity for its ability to adapt pretrained models to target domains without accessing source domains, ensuring source data privacy. While SFUDA is well-developed in visual tasks, its application to Time-Series SFUD... | [
{
"id": "lXb2poNAol",
"initial_rating": 6,
"confidence": 3,
"soundness": 2,
"contribution": 2,
"presentation": 3,
"summary": "This paper addresses the time-series source-free domain adaptation problem. It proposes a source recovery method that recovers target data to a source-like distri... | {
"rating": "3;5;6",
"rating_avg": 4.666666666666667,
"confidence": "3;5;3",
"confidence_avg": 3.6666666666666665,
"soundness": "3;2;2",
"soundness_avg": 2.3333333333333335,
"contribution": "2;3;2",
"contribution_avg": 2.3333333333333335,
"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": "2026-01-14T22:16:01.465131"
} | {
"id": "G96UGpbr63",
"metareview": "The paper introduces a method for source distribution recovery with a detailed methodological presentation and clearly defined variants. The idea of source distribution recovery is well-motivated, and the method section is clearly and comprehensively presented. The ablation stud... | {
"decision": "Reject"
} |
GhexuBLxbO | 2410.08815v2 | StructRAG: Boosting Knowledge Intensive Reasoning of LLMs via Inference-time Hybrid Information Structurization | {
"content": "## Abstract\n\nAbstract Retrieval-augmented generation (RAG) is a key means to effectively enhance large language models (LLMs) in many knowledge-based tasks.\nHowever, existing RAG methods struggle with knowledge-intensive reasoning tasks, because useful information required to these tasks are badly sc... | [
{
"id": "Q3n3ZTdaRE",
"initial_rating": 6,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "The paper introduces StructRAG, a novel framework designed to enhance LLMs for knowledge-intensive reasoning tasks by structuring information at inference time. S... | {
"rating": "5;6;8;8",
"rating_avg": 6.75,
"confidence": "3;4;4;4",
"confidence_avg": 3.75,
"soundness": "3;3;4;3",
"soundness_avg": 3.25,
"contribution": "3;3;3;3",
"contribution_avg": 3,
"presentation": "3;3;4;3",
"presentation_avg": 3.25
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.465923"
} | {
"id": "w1xC6uZ0uo",
"metareview": "This paper proposes a novel framework for solving knowledge-intensive reasoning tasks, which are difficult for existing retrieval-augmentation generation methods to address in locating key information. The framework comprises three components: a structure router for structure ty... | {
"decision": "Accept (Poster)"
} |
Gi3SwL98nL | 2410.11522v2 | Leveraging LLM Embeddings for Cross Dataset Label Alignment and Zero Shot Music Emotion Prediction | {
"content": "## Abstract\n\nAbstract In this work, we present a novel method for music emotion recognition that leverages Large Language Model (LLM) embeddings for label alignment across multiple datasets and zero-shot prediction on novel categories. First, we compute LLM embeddings for emotion labels and apply non-... | [
{
"id": "CMv8XK68G9",
"initial_rating": 3,
"confidence": 3,
"soundness": 3,
"contribution": 2,
"presentation": 2,
"summary": "The authors use sentence embedding (from Sentence-BERT) and MERT to get the embedding of music emotion label and music respectively, and try to align the label an... | {
"rating": "3;3;5;5",
"rating_avg": 4,
"confidence": "5;3;3;4",
"confidence_avg": 3.75,
"soundness": "2;3;3;3",
"soundness_avg": 2.75,
"contribution": "2;2;2;3",
"contribution_avg": 2.25,
"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": "2026-01-14T22:16:01.466525"
} | {
"id": "hs0cXWcAuQ",
"metareview": "The paper introduced a method to conduct zero-shot emotion recognition from music audio across multiple datasets. The authors follow the well-known zero-shot learning framework and extract a language embedding from the label text using SentenceBERT (which they call an \"LLM\"), ... | {
"decision": "Reject"
} |
GiHLTtfbB5 | 2311.01698v1 | Adversarial Attacks on Cooperative Multi-agent Bandits | {
"content": "## Abstract\n\nAbstract Cooperative multi-agent multi-armed bandits ( CMA2B ) consider the collaborative efforts of multiple agents in a shared multi-armed bandit game. We study latent vulnerabilities exposed by this collaboration and consider adversarial attacks on a few agents with the goal of influen... | [
{
"id": "ClSSBJe1VU",
"initial_rating": 6,
"confidence": 4,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "This paper studies adversarial attack strategies in the Cooperative Multi-Arm Bandit setting CMA2B. In the Cooperative Multi-Arm bandit setting ($M$) agents pull ... | {
"rating": "3;3;5;5;6",
"rating_avg": 4.4,
"confidence": "4;5;3;3;4",
"confidence_avg": 3.8,
"soundness": "1;2;3;3;3",
"soundness_avg": 2.4,
"contribution": "2;1;2;2;2",
"contribution_avg": 1.8,
"presentation": "1;3;3;1;3",
"presentation_avg": 2.2
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:01.467249"
} | {
"id": "38rF6TjuJ6",
"metareview": "This paper studies adversarial attacks on cooperative multi-agent multi-armed bandits (CMA2B) in both homogeneous and heterogeneous settings. The reviewers agree that the paper addresses an interesting and well-motivated problem. However, the reviewers raised several shared conc... | {
"decision": "Reject"
} |
Gj5JTAwdoy | 2410.05167v1 | Presto! Distilling Steps and Layers for Accelerating Music Generation | {
"content": "## Abstract\n\nAbstract Despite advances in diffusion-based text-to-music (TTM) methods, efficient, high-quality generation remains a challenge.\nWe introduce Presto! , an approach to inference acceleration for score-based diffusion transformers via reducing both sampling steps and cost per step.\nTo re... | [
{
"id": "Z9NflfrKww",
"initial_rating": 8,
"confidence": 4,
"soundness": 4,
"contribution": 4,
"presentation": 4,
"summary": "The paper proposes \"Presto!\" for effective and efficient music generation. More specificly, Presto! is a set of model distillation techniques that aims at impro... | {
"rating": "5;8;8;8",
"rating_avg": 7.25,
"confidence": "4;3;4;4",
"confidence_avg": 3.75,
"soundness": "3;4;4;4",
"soundness_avg": 3.75,
"contribution": "2;3;3;4",
"contribution_avg": 3,
"presentation": "2;3;3;4",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Spotlight",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.468130"
} | {
"id": "1Xl263NmAf",
"metareview": "**Paper Summary:**\n\nThis paper improves the inference speed of text-to-music diffusion models, adapting techniques for model distillation (DMD2) and inference acceleration (ASE) from prior work on vision. These adaptations are well-motivated, and experiments convincingly valid... | {
"decision": "Accept (Spotlight)"
} |
GjM61KRiTG | 2408.15313v1 | Bi-Factorial Preference Optimization: Balancing Safety-Helpfulness in Language Models | {
"content": "## Abstract\n\nAbstract Fine-tuning large language models (LLMs) on human preferences, typically through reinforcement learning from human feedback (RLHF), has proven successful in enhancing their capabilities. However, ensuring the safety of LLMs during the fine-tuning remains a critical concern, and m... | [
{
"id": "SoifO3jl2M",
"initial_rating": 6,
"confidence": 3,
"soundness": 3,
"contribution": 2,
"presentation": 2,
"summary": "This paper proposes Bi-Factorial Preference Optimization (BFPO) to address the limitations in balancing the helpfulness and safety. Specifically, this paper conve... | {
"rating": "5;6;8;8",
"rating_avg": 6.75,
"confidence": "3;3;3;4",
"confidence_avg": 3.25,
"soundness": "3;3;4;3",
"soundness_avg": 3.25,
"contribution": "3;2;3;3",
"contribution_avg": 2.75,
"presentation": "2;2;3;3",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Spotlight",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.468951"
} | {
"id": "NDER8OQVnL",
"metareview": "This paper proposes a novel loss function to balance helpfulness and safety in large language model alignment by transforming multi-objective RLHF into a direct preference-based loss. Experiments on various alignment datasets, including those focused on helpfulness and safety, d... | {
"decision": "Accept (Spotlight)"
} |
GjSstLcxAs | 2410.00359v1 | Self-controller: Controlling LLMs with Multi-round Step-by-step Self-awareness | {
"content": "## Abstract\n\nAbstract The applications of large language models (LLMs) have been widely spread across all domains.\nHowever, the basic abilities such as the controllability of LLMs are still limited.\nTo address this, we propose ” Self-controller ”, a novel agentic framework bringing self-awareness in... | [
{
"id": "oTAOF2zNxU",
"initial_rating": 3,
"confidence": 4,
"soundness": 2,
"contribution": 2,
"presentation": 2,
"summary": "The authors present Self-controller, a multi-round prompting approach to more efficiently guide LLMs to generate text with length constraints. The authors claim t... | {
"rating": "3;3;3;5",
"rating_avg": 3.5,
"confidence": "3;3;4;3",
"confidence_avg": 3.25,
"soundness": "2;3;2;3",
"soundness_avg": 2.5,
"contribution": "2;2;2;2",
"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": "2026-01-14T22:16:01.469741"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
GkJCgUmIqA | 2409.10777v1 | Physics-Informed Neural Networks with Trust-Region Sequential Quadratic Programming | {
"content": "## Abstract\n\nAbstract Physics-Informed Neural Networks (PINNs) represent a significant advancement in Scientific Machine Learning (SciML), which integrate physical domain knowledge into an empirical loss function as soft constraints and apply existing machine learning methods to train the model.\nHowe... | [
{
"id": "JrodpOUC6W",
"initial_rating": 3,
"confidence": 3,
"soundness": 2,
"contribution": 2,
"presentation": 2,
"summary": "This paper focuses on the optimization problems in PINN methods. The authors introduce the trust-region sequential quadratic programming to solve the constraint i... | {
"rating": "3;3;3",
"rating_avg": 3,
"confidence": "3;4;3",
"confidence_avg": 3.3333333333333335,
"soundness": "2;2;2",
"soundness_avg": 2,
"contribution": "2;2;2",
"contribution_avg": 2,
"presentation": "4;3;2",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:01.470480"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
GkWA6NjePN | 2410.18636v1 | Multi-agent cooperation through learning-aware policy gradients | {
"content": "## Abstract\n\nAbstract Self-interested individuals often fail to cooperate, posing a fundamental challenge for multi-agent learning. How can we achieve cooperation among self-interested, independent learning agents? Promising recent work has shown that in certain tasks cooperation can be established be... | [
{
"id": "PIeO0byRW3",
"initial_rating": 6,
"confidence": 4,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "This paper introduces a new learning-aware reinforcement learning rule derived as a policy gradient estimator. Instead of using long histories of samples, the pol... | {
"rating": "3;5;5;5;6;8",
"rating_avg": 5.333333333333333,
"confidence": "4;4;4;4;4;2",
"confidence_avg": 3.6666666666666665,
"soundness": "3;3;2;2;3;3",
"soundness_avg": 2.6666666666666665,
"contribution": "3;2;3;2;2;3",
"contribution_avg": 2.5,
"presentation": "1;3;4;2;3;4",
"presentation_avg": 2... | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.471579"
} | {
"id": "skMwNZ6pbG",
"metareview": "The reviewers acknowledged that the paper studies an interesting problem of cooperation among self-interested learning agents in multi-agent systems, and the proposed method of learning-aware policy gradients is novel with significant contributions. However, the reviewers also ... | {
"decision": "Accept (Poster)"
} |
Gl2nXRzclw | 2301.07858v1 | Robust Gaussian Process Regression with Huber Likelihood | {
"content": "## Abstract\n\nAbstract Gaussian process regression in its most simplified form assumes normal homoscedastic noise and utilizes analytically tractable mean and covariance functions of predictive posterior distribution using Gaussian conditioning. Its hyperparameters are estimated by maximizing the evide... | [
{
"id": "PBeoUPh7KW",
"initial_rating": 5,
"confidence": 3,
"soundness": 2,
"contribution": 2,
"presentation": 2,
"summary": "This paper presents a novel method for robust gaussian process regression by using huber likelihood, which can deal with outliers in both covariate and output spa... | {
"rating": "3;5;5;6",
"rating_avg": 4.75,
"confidence": "5;5;3;4",
"confidence_avg": 4.25,
"soundness": "2;3;2;3",
"soundness_avg": 2.5,
"contribution": "2;2;2;3",
"contribution_avg": 2.25,
"presentation": "1;3;2;4",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:01.472409"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
GlPVnuL66V | 2410.07632v1 | Provable Privacy Attacks on Trained Shallow Neural Networks | {
"content": "## Abstract\n\nAbstract We study what provable privacy attacks can be shown on trained, 2-layer ReLU neural networks. We explore two types of attacks; data reconstruction attacks, and membership inference attacks. We prove that theoretical results on the implicit bias of 2-layer neural networks can be u... | [
{
"id": "lp77TCWCg9",
"initial_rating": 6,
"confidence": 3,
"soundness": 2,
"contribution": 3,
"presentation": 1,
"summary": "The paper tries to show provable guarantees for data reconstruction and membership inference for nonconvex activation functions in shallow Neural networks.",
... | {
"rating": "3;3;5;8",
"rating_avg": 4.75,
"confidence": "4;3;3;3",
"confidence_avg": 3.25,
"soundness": "2;1;2;3",
"soundness_avg": 2,
"contribution": "1;2;2;4",
"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": "2026-01-14T22:16:01.473310"
} | {
"id": "7YEHeP8OdU",
"metareview": "This paper studies the privacy attack on 2-layer Relu activated neural network. Under the assumption that the model converges to a point that satisfies the KKT condition, this paper shows 2 attacks: 1) for the univariate case, this paper is able to recover a part of the training... | {
"decision": "Reject"
} |
Glm7Kj47nN | 2409.03718v1 | Geometry Image Diffusion: Fast and Data-Efficient Text-to-3D with Image-Based Surface Representation | {
"content": "## Abstract\n\nAbstract Generating high-quality 3D objects from textual descriptions remains a challenging problem due to computational cost, the scarcity of 3D data, and complex 3D representations.\nWe introduce Geometry Image Diffusion (GIMDiffusion), a novel Text-to-3D model that utilizes geometry im... | [
{
"id": "lKbzYJto4Q",
"initial_rating": 8,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "This paper studies how to generate high quality mesh of objects with GIM.\n\nThe 3D objects are parameterized into 2D chart geometry image. This representation is... | {
"rating": "5;5;5;8",
"rating_avg": 5.75,
"confidence": "4;4;4;3",
"confidence_avg": 3.75,
"soundness": "3;3;3;3",
"soundness_avg": 3,
"contribution": "3;2;2;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": "2026-01-14T22:16:01.474225"
} | {
"id": "AM7hpjGCTW",
"metareview": "This paper introduces Geometry Image Diffusion, a novel framework for Text-to-3D generation that leverages 2D diffusion-based techniques to represent 3D shapes through geometry images. By sidestepping the complexity of existing 3D-aware architectures, the method significantly en... | {
"decision": "Accept (Poster)"
} |
GpdO9r73xT | 2406.01970v1 | The Crystal Ball Hypothesis in diffusion models: Anticipating object positions from initial noise | {
"content": "## Abstract\n\nAbstract Diffusion models have achieved remarkable success in text-to-image generation tasks; however, the role of initial noise has been rarely explored. In this study, we identify specific regions within the initial noise image, termed trigger patches, that play a key role for object ge... | [
{
"id": "Zs4oGkE8MU",
"initial_rating": 8,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "In the paper, the authors introduce the concept of \"trigger patches,\" which are specific regions within the initial noise image that determine the positional in... | {
"rating": "5;6;6;8",
"rating_avg": 6.25,
"confidence": "3;4;4;3",
"confidence_avg": 3.5,
"soundness": "3;3;3;3",
"soundness_avg": 3,
"contribution": "3;3;3;3",
"contribution_avg": 3,
"presentation": "3;3;3;3",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.474871"
} | {
"id": "8WtR8goGWi",
"metareview": "This paper explores the influence of initial noise in the text-to-image generation task, and proposes the concept of triggle patches that play a key role in inducing object generation in the resulting images. All reviewers agree the insight of triggle patches are interesting, an... | {
"decision": "Accept (Poster)"
} |
Gq7RDMeZi4 | 2310.12457v1 | Forming Scalable, Convergent GNN Layers that Minimize a Sampling-Based Energy | {
"content": "## Abstract\n\nAbstract This document is a model and instructions for L a T e X .\nThis and the IEEEtran.cls file define the components of your paper [title, text, heads, etc.]. *CRITICAL: Do Not Use Symbols, Special Characters, Footnotes,\nor Math in Paper Title or Abstract.\n\n###### Index Terms:\n\n#... | [
{
"id": "cQkqP9yrSw",
"initial_rating": 8,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "The paper proposes a GNN model that scales effectively to large datasets by incorporating sampled subgraphs into its energy function design. This approach allows... | {
"rating": "5;6;6;8",
"rating_avg": 6.25,
"confidence": "4;3;3;3",
"confidence_avg": 3.25,
"soundness": "3;3;3;3",
"soundness_avg": 3,
"contribution": "3;3;3;3",
"contribution_avg": 3,
"presentation": "2;3;3;3",
"presentation_avg": 2.75
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.475479"
} | {
"id": "TAGZlZRmse",
"metareview": "The manuscript proposes a graphical neural network training strategy based on minimizing a sub-sampled graphs into the energy minimization. Empirical results suggest the effectiveness and scalability of the training strategy. The reviewers largely agree that the approach is nove... | {
"decision": "Accept (Poster)"
} |
GqBO71SPjL | 2410.14462v1 | LUDVIG: Learning-free Uplifting of 2D Visual Features to Gaussian Splatting Scenes | {
"content": "## Abstract\n\nAbstract We address the task of uplifting visual features or semantic masks from 2D vision models to 3D scenes represented by Gaussian Splatting. Whereas common approaches rely on iterative optimization-based procedures, we show that a simple yet effective aggregation technique yields exc... | [
{
"id": "xN9pWyvREi",
"initial_rating": 3,
"confidence": 3,
"soundness": 2,
"contribution": 2,
"presentation": 1,
"summary": "This paper is primarily concerned with generating 3D feature representations of a scene given multiple images of a stationary scene captured from different viewpo... | {
"rating": "3;3;6;6",
"rating_avg": 4.5,
"confidence": "2;3;3;3",
"confidence_avg": 2.75,
"soundness": "2;2;3;3",
"soundness_avg": 2.5,
"contribution": "1;2;3;3",
"contribution_avg": 2.25,
"presentation": "3;1;3;3",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:01.475976"
} | {
"id": "3QYzCMwjcg",
"metareview": "This paper proposes a method that lift 2D features from foundational models (e.g., CLIP, SAM) to learned 3D Gaussians for a scene. The main idea is to apply weight aggregation on the features of each pixel, as well as a diffusion mechanism that diffuse features to nearby Gaussia... | {
"decision": "Reject"
} |
GqhvJ1o8m5 | 2410.00262v1 | ImmersePro: End-to-End Stereo Video Synthesis Via Implicit Disparity Learning | {
"content": "## Abstract\n\nAbstract We introduce ImmersePro , an innovative framework specifically designed to transform single-view videos into stereo videos. This framework utilizes a novel dual-branch architecture comprising a disparity branch and a context branch on video data by leveraging spatial-temporal att... | [
{
"id": "KnbuxLxeXq",
"initial_rating": 6,
"confidence": 5,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "This paper introduces a method to convert single-view videos into stereo videos. The authors propose a dual-branch architecture with spatial-temporal attention la... | {
"rating": "3;3;3;6",
"rating_avg": 3.75,
"confidence": "5;4;3;5",
"confidence_avg": 4.25,
"soundness": "2;2;2;3",
"soundness_avg": 2.25,
"contribution": "2;2;2;3",
"contribution_avg": 2.25,
"presentation": "1;2;2;3",
"presentation_avg": 2
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:01.476672"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
Gr8nHvOivO | 2311.01479v6 | Detecting Out-of-Distribution through the Lens of Neural Collapse | {
"content": "## Abstract\n\nAbstract Out-of-Distribution (OOD) detection is essential for safe deployment; however, existing detectors exhibit generalization discrepancies and cost concerns.\nTo address this, we propose a highly versatile and efficient OOD detector inspired by the trend of Neural Collapse on practic... | [
{
"id": "bBm5O2v7SQ",
"initial_rating": 3,
"confidence": 4,
"soundness": 2,
"contribution": 2,
"presentation": 3,
"summary": "The paper presents an approach to Out-of-Distribution (OOD) detection in machine learning models, leveraging the concept of Neural Collapse. The authors argue tha... | {
"rating": "3;5;5;5",
"rating_avg": 4.5,
"confidence": "4;4;4;4",
"confidence_avg": 4,
"soundness": "2;3;2;3",
"soundness_avg": 2.5,
"contribution": "2;3;2;1",
"contribution_avg": 2,
"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": "2026-01-14T22:16:01.477504"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
GrDne4055L | 2410.10744v1 | Adversarially Robust Out-of-Distribution Detection Using Lyapunov-Stabilized Embeddings | {
"content": "## Abstract\n\nAbstract Despite significant advancements in out-of-distribution (OOD) detection, existing methods still struggle to maintain robustness against adversarial attacks, compromising their reliability in critical real-world applications. Previous studies have attempted to address this challen... | [
{
"id": "wKnbtHnr5t",
"initial_rating": 8,
"confidence": 2,
"soundness": 4,
"contribution": 3,
"presentation": 3,
"summary": "This paper proposes a new out-of-distribution detector that is robust against adversarial attacks. Specifically, the authors point out the problems of adversarial... | {
"rating": "5;6;6;8",
"rating_avg": 6.25,
"confidence": "4;3;4;2",
"confidence_avg": 3.25,
"soundness": "2;3;2;4",
"soundness_avg": 2.75,
"contribution": "3;3;3;3",
"contribution_avg": 3,
"presentation": "3;3;3;3",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.478358"
} | {
"id": "qUvvN0NgOb",
"metareview": "Out-of-distribution detection aims to improve the systems' robustness by letting the system say no to something unknown to it. However, the safety of OOD detection algorithm is rarely discussed or studied. This paper focuses on this important problem and makes a further step tow... | {
"decision": "Accept (Poster)"
} |
GrHewano8m | 2406.12693v1 | XXLTraffic: Expanding and Extremely Long Traffic forecasting beyond test adaptation | {
"content": "## Abstract\n\nAbstract Traffic forecasting is crucial for smart cities and intelligent transportation initiatives, where deep learning has made significant progress in modeling complex spatio-temporal patterns in recent years. However, current public datasets have limitations in reflecting the ultra-dy... | [
{
"id": "6oTT07OEV4",
"initial_rating": 5,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 2,
"summary": "The authors present XXLTraffic, the longest available public traffic dataset with the longest timespan collected from Los Angeles, USA, and New South Wales, Austr... | {
"rating": "3;5;5;8",
"rating_avg": 5.25,
"confidence": "4;5;4;4",
"confidence_avg": 4.25,
"soundness": "1;2;3;4",
"soundness_avg": 2.5,
"contribution": "2;2;3;4",
"contribution_avg": 2.75,
"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": "2026-01-14T22:16:01.479056"
} | {
"id": "ZWg5On8edv",
"metareview": "This paper presents XXLTraffic with the longest timespan collected from Los Angeles, USA, and New South Wales, Australia, curated to support research in extremely long forecasting beyond test adaptation. The reviewers and I believe that providing this benchmark is a very meaning... | {
"decision": "Reject"
} |
GraOHDxFjl | 2404.06773v4 | LLaMA Decoder As Vision Transformer | {
"content": "## Abstract\n\nAbstract This work examines whether decoder-only Transformers such as LLaMA, which were originally designed for large language models (LLMs), can be adapted to the computer vision field.\nWe first \"LLaMAfy\" a standard ViT step-by-step to align with LLaMA’s architecture, and find that di... | [
{
"id": "SxRBrKSD8o",
"initial_rating": 3,
"confidence": 4,
"soundness": 3,
"contribution": 1,
"presentation": 3,
"summary": "The paper investigates the potential of using a decoder-only Transformer architecture, specifically LLaMa, as a vision Transformer (ViT) classifier for image proc... | {
"rating": "3;3;6;8",
"rating_avg": 5,
"confidence": "5;4;4;4",
"confidence_avg": 4.25,
"soundness": "2;3;3;3",
"soundness_avg": 2.75,
"contribution": "1;1;3;3",
"contribution_avg": 2,
"presentation": "2;3;4;4",
"presentation_avg": 3.25
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:01.479730"
} | {
"id": "biFwzAfxmM",
"metareview": "This paper presents an interesting idea that makes use of the architecture of the LLaMA model (a language model) as a vision transformer. For this purpose, the LLaMA architecture is step-by-step adjusted to fit the vision data (see Figure 1, the right-hand side). While the final... | {
"decision": "Reject"
} |
GrmFFxGnOR | 2410.01201v2 | Were RNNs All We Needed? | {
"content": "## Abstract\n\nAbstract The scalability limitations of Transformers regarding sequence length have renewed interest in recurrent sequence models that are parallelizable during training. As a result, many novel recurrent architectures, such as S4, Mamba, and Aaren, have been proposed that achieve compara... | [
{
"id": "elEWy4ZKtn",
"initial_rating": 3,
"confidence": 4,
"soundness": 4,
"contribution": 4,
"presentation": 3,
"summary": "The paper introduces minimal Long Short-Term Memory (minLSTM) and minimal Gated Recurrent Unit (minGRU) models.\nThe authors modify the traditional LSTM and GRU m... | {
"rating": "3;3;5;8",
"rating_avg": 4.75,
"confidence": "4;4;5;3",
"confidence_avg": 4,
"soundness": "4;3;3;4",
"soundness_avg": 3.5,
"contribution": "1;1;2;4",
"contribution_avg": 2,
"presentation": "4;3;3;3",
"presentation_avg": 3.25
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:01.480424"
} | {
"id": "Q1p50Gtsqm",
"metareview": "This paper introduces minLSTM and minGRU, simplified variants of traditional LSTM and GRU models, designed to enable parallel training by removing hidden state dependencies. These modifications result in models that are computationally efficient and achieve competitive performan... | {
"decision": "Reject"
} |
GsGmdxcFNL | 2410.14081v1 | Reward-free World Models for Online Imitation Learning | {
"content": "## Abstract\n\nAbstract Imitation learning (IL) enables agents to acquire skills directly from expert demonstrations, providing a compelling alternative to reinforcement learning. However, prior online IL approaches struggle with complex tasks characterized by high-dimensional inputs and complex dynamic... | [
{
"id": "GfPO5NROG2",
"initial_rating": 3,
"confidence": 5,
"soundness": 1,
"contribution": 2,
"presentation": 1,
"summary": "This paper introduces an extension of the TD-MPC framework for imitation, specifically within the inverse reinforcement learning context. Rather than learning an ... | {
"rating": "3;5;5;6",
"rating_avg": 4.75,
"confidence": "5;4;5;1",
"confidence_avg": 3.75,
"soundness": "1;3;2;3",
"soundness_avg": 2.25,
"contribution": "2;2;2;3",
"contribution_avg": 2.25,
"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": "2026-01-14T22:16:01.481132"
} | {
"id": "V4BamlzfEP",
"metareview": "The paper proposes IQ-MPC, a model-based, reward-free imitation learning framework that combines inverse soft-Q learning (IQ-Learn) with model predictive control (MPC). Instead of explicitly learning reward functions, IQ-MPC operates in the Q-function space and employs latent dy... | {
"decision": "Reject"
} |
GtvuNrk58a | 2410.06205v1 | Round and Round We Go! What makes Rotary Positional Encodings useful? | {
"content": "## Abstract\n\nAbstract Positional Encodings (PEs) are a critical component of Transformer-based Large Language Models (LLMs), providing the attention mechanism with important sequence-position information. One of the most popular types of encoding used today in LLMs are Rotary Positional Encodings (RoP... | [
{
"id": "raH0YWgspE",
"initial_rating": 8,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "This paper delves into the role of Position Embedding (PE) in LLMs, challenging the traditional view that RoPE primarily attenuates attention weights as the relat... | {
"rating": "3;5;5;5;8",
"rating_avg": 5.2,
"confidence": "3;4;4;4;4",
"confidence_avg": 3.8,
"soundness": "2;2;2;2;3",
"soundness_avg": 2.2,
"contribution": "2;1;3;2;4",
"contribution_avg": 2.4,
"presentation": "3;3;3;3;3",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.481888"
} | {
"id": "iSH3ZBnZXn",
"metareview": "**Summary:** This paper provides a theoretical and empirical study of Rotary Positional Encodings (RoPE) in transformer-based LLMs. The authors challenge the conventional belief that RoPE facilitates token dependency decay with increasing distance. Instead, the work hypothesizes... | {
"decision": "Accept (Poster)"
} |
Gv0TOAigIY | 2408.15495v1 | Remove Symmetries to Control Model Expressivity | {
"content": "## Abstract\n\nAbstract When symmetry is present in the loss function, the model is likely to be trapped in a low-capacity state that is sometimes known as a “collapse.” Being trapped in these low-capacity states can be a major obstacle to training across many scenarios where deep learning technology is... | [
{
"id": "KjhJ9M4T0W",
"initial_rating": 6,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "This paper presents a novel approach for addressing the limitations posed by symmetries in neural network training, which often lead to low-capacity model states.... | {
"rating": "5;5;6;6",
"rating_avg": 5.5,
"confidence": "3;3;2;3",
"confidence_avg": 2.75,
"soundness": "3;3;3;3",
"soundness_avg": 3,
"contribution": "2;2;3;3",
"contribution_avg": 2.5,
"presentation": "3;2;3;3",
"presentation_avg": 2.75
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.482747"
} | {
"id": "5w5dnEGbMv",
"metareview": "**Summary** This paper studies the role of neural network parameter symmetries on the phenomenon of collapse in which the model is trapped in a low-capacity state during training. The authors provide theory explaining how symmetry can lead to reduced capacity. That propose a m... | {
"decision": "Accept (Poster)"
} |
GvUahyZ8UF | 2410.04461v1 | Improved Off-policy Reinforcement Learning in Biological Sequence Design | {
"content": "## Abstract\n\nAbstract Designing biological sequences with desired properties is a significant challenge due to the combinatorially vast search space and the high cost of evaluating each candidate sequence.\nTo address these challenges, reinforcement learning (RL) methods, such as GFlowNets, utilize pr... | [
{
"id": "CEpQuSjrI6",
"initial_rating": 8,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "The paper studies the problem of generative search, where we seek to generate novel and diverse particles from a large design space that yield high reward. A core... | {
"rating": "3;3;5;8",
"rating_avg": 4.75,
"confidence": "4;1;3;3",
"confidence_avg": 2.75,
"soundness": "2;2;3;3",
"soundness_avg": 2.5,
"contribution": "2;2;2;3",
"contribution_avg": 2.25,
"presentation": "3;3;2;3",
"presentation_avg": 2.75
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:01.483487"
} | {
"id": "WT1dSillCm",
"metareview": "This paper propose a new robust RL method to combat reward misspecification focusing on the biological sequence design task. There are a lot of back and forth discussion between the author and reviewers. \n\nThe main drawback of the paper lies in its inadequate experimental eval... | {
"decision": "Reject"
} |
GwSL33Qx42 | 2410.13370v1 | MagicTailor: Component-Controllable Personalization in Text-to-Image Diffusion Models | {
"content": "## Abstract\n\nAbstract Recent advancements in text-to-image (T2I) diffusion models have enabled the creation of high-quality images from text prompts, but they still struggle to generate images with precise control over specific visual concepts. Existing approaches can replicate a given concept by lear... | [
{
"id": "AGVsGmYgEE",
"initial_rating": 5,
"confidence": 4,
"soundness": 2,
"contribution": 3,
"presentation": 2,
"summary": "This paper proposes a new architecture for component-controllable personalization. It introduces two main design choices: Dynamic Masked Degradation (DM-Deg) to p... | {
"rating": "3;5;5",
"rating_avg": 4.333333333333333,
"confidence": "4;4;4",
"confidence_avg": 4,
"soundness": "2;3;2",
"soundness_avg": 2.3333333333333335,
"contribution": "2;2;3",
"contribution_avg": 2.3333333333333335,
"presentation": "3;3;2",
"presentation_avg": 2.6666666666666665
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:01.484282"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
Gws8Q4wSrJ | 2304.03641v2 | A Block Coordinate Descent Method for Nonsmooth Composite Optimization under Orthogonality Constraints | {
"content": "## Abstract\n\nAbstract Nonsmooth composite optimization with orthogonality constraints is crucial in statistical learning and data science, but it presents challenges due to its nonsmooth objective and computationally expensive, non-convex constraints. In this paper, we propose a new approach called OB... | [
{
"id": "eZRh4Bcpxz",
"initial_rating": 3,
"confidence": 3,
"soundness": 3,
"contribution": 2,
"presentation": 1,
"summary": "The authors study a the problem of minimizing a particular class nonconvex nonsmooth function $f$ over the space of n x r orthogonal matrices (a Stiefel manifold ... | {
"rating": "3;3;6",
"rating_avg": 4,
"confidence": "4;3;4",
"confidence_avg": 3.6666666666666665,
"soundness": "2;3;3",
"soundness_avg": 2.6666666666666665,
"contribution": "2;2;3",
"contribution_avg": 2.3333333333333335,
"presentation": "2;1;2",
"presentation_avg": 1.6666666666666667
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:01.485463"
} | {
"id": "H7jAhKWpyX",
"metareview": "This paper studies block coordinate descent method (BCD) for nonsmooth composite optimization with orthogonality constraints. The authors proposed a new method that minimizes a majorizing surrogate of the block restricted objective, which is obtained by sampling k rows and formu... | {
"decision": "Reject"
} |
GzLepH6MBB | 2405.00448v2 | MMTryon: Multi-Modal Multi-Reference Control for High-Quality Fashion Generation | {
"content": "## Abstract\n\nAbstract This paper introduces MMTryon, a multi-modal multi-reference VIrtual Try-ON (VITON) framework, which can generate high-quality compositional try-on results by taking a text instruction and multiple garment images as inputs. Our MMTryon addresses three problems overlooked in prior... | [
{
"id": "tIDF1B69yz",
"initial_rating": 8,
"confidence": 5,
"soundness": 4,
"contribution": 4,
"presentation": 4,
"summary": "The paper introduces MMTryon, a model that integrates a multi-modality and multi-reference attention mechanism to enhance virtual try-on tasks by combining garmen... | {
"rating": "3;8;8",
"rating_avg": 6.333333333333333,
"confidence": "5;5;5",
"confidence_avg": 5,
"soundness": "2;4;4",
"soundness_avg": 3.3333333333333335,
"contribution": "2;4;4",
"contribution_avg": 3.3333333333333335,
"presentation": "2;4;4",
"presentation_avg": 3.3333333333333335
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:01.486890"
} | {
"id": "tCG8cAizRu",
"metareview": "The paper receives mixed scores from the reviewers. While the reviewers appreciate the innovative design and extensive experiments, they also raised some concerns including claims overstated, the contributions incremental, and the texture quality limited.\n\nIn particular, M&M V... | {
"decision": "Reject"
} |
H0qIWXXLUR | 2404.09656v3 | Learn Your Reference Model for Real Good Alignment | {
"content": "## Abstract\n\nAbstract Despite the fact that offline methods for Large Language Models (LLMs) alignment do not require a direct reward model, they remain susceptible to overoptimization. This issue arises when the trained model deviates excessively from the reference policy, leading to a decrease in sa... | [
{
"id": "kbFMFkiljO",
"initial_rating": 6,
"confidence": 4,
"soundness": 3,
"contribution": 2,
"presentation": 4,
"summary": "Recent literature on Offline RL LM fine-tuning methods, such as DPO, IPO, KTO etc, involve avoiding training an explicitly reward model (RM) and directly optimizi... | {
"rating": "3;5;5;5;6",
"rating_avg": 4.8,
"confidence": "4;4;4;5;4",
"confidence_avg": 4.2,
"soundness": "3;3;3;2;3",
"soundness_avg": 2.8,
"contribution": "2;3;2;2;2",
"contribution_avg": 2.2,
"presentation": "2;2;3;3;4",
"presentation_avg": 2.8
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.487609"
} | {
"id": "4JRuZJsGMy",
"metareview": "This paper aims to mitigate the overoptimization problem in offline RL methods for aligning language models. The proposed method brings in idea from trust region optimization and is based on the idea of updating the reference language model to make it a moving target. Experiment... | {
"decision": "Accept (Poster)"
} |
H25xduunIK | 2409.00844v1 | Report Cards: Qualitative Evaluation of Language Models Using Natural Language Summaries | {
"content": "## Abstract\n\nAbstract The rapid development and dynamic nature of large language models (LLMs) make it difficult for conventional quantitative benchmarks to accurately assess their capabilities.\nWe propose Report Cards, which are human-interpretable, natural language summaries of model behavior for s... | [
{
"id": "13st3up54m",
"initial_rating": 5,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "This paper proposes a framework for qualitative evaluation, based on specificity, faithfulness, and interoperability. To do so, the paper proposes a PRESS algorit... | {
"rating": "3;5;6;6",
"rating_avg": 5,
"confidence": "4;4;4;3",
"confidence_avg": 3.75,
"soundness": "2;2;3;3",
"soundness_avg": 2.5,
"contribution": "1;2;3;3",
"contribution_avg": 2.25,
"presentation": "3;4;4;3",
"presentation_avg": 3.5
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:01.488415"
} | {
"id": "VOePim0kfh",
"metareview": "The paper introduces \"Report Cards\" as a qualitative evaluation and description of large language models through human-interpretable natural language summaries. The authors proposed methods to evaluate the quality of generated cards. They also proposed an iterative algorithm c... | {
"decision": "Reject"
} |
H2Gxil855b | 2408.13055v2 | Atlas Gaussians Diffusion for 3D Generation | {
"content": "## Abstract\n\nAbstract Using the latent diffusion model has proven effective in developing novel 3D generation techniques. To harness the latent diffusion model, a key challenge is designing a high-fidelity and efficient representation that links the latent space and the 3D space.\nIn this paper, we in... | [
{
"id": "KQiQLXO58Z",
"initial_rating": 8,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 4,
"summary": "This paper introduces Atlas Gaussians, a new latent representation learning framework designed to improve the efficiency of 3D object generation by addressing the... | {
"rating": "5;6;8;8",
"rating_avg": 6.75,
"confidence": "3;4;3;3",
"confidence_avg": 3.25,
"soundness": "3;3;3;3",
"soundness_avg": 3,
"contribution": "2;3;3;3",
"contribution_avg": 2.75,
"presentation": "3;2;4;4",
"presentation_avg": 3.25
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Spotlight",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.489158"
} | {
"id": "awaj3q1iCa",
"metareview": "This paper proposes a novel framework, Atlas Gaussians Diffusion, for 3D shape generation. The method represents 3D shapes as a union of local patches defined as 3D Gaussians in UV space, enabling the generation of theoretically unlimited points while preserving finer details. T... | {
"decision": "Accept (Spotlight)"
} |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.