paper_id string | arxiv_id string | title string | markdown dict | reviews list | scores dict | metadata dict | meta_review dict | decision dict |
|---|---|---|---|---|---|---|---|---|
4S9bBbX1be | 2408.00415v1 | DriveArena: A Closed-loop Generative Simulation Platform for Autonomous Driving | {
"content": "## Abstract\n\nAbstract This paper presented DriveArena , the first high-fidelity closed-loop simulation system designed for driving agents navigating in real scenarios. DriveArena features a flexible, modular architecture, allowing for the seamless interchange of its core components: Traffic Manager, a... | [
{
"id": "2Fy2YBEh7B",
"initial_rating": 3,
"confidence": 4,
"soundness": 2,
"contribution": 2,
"presentation": 2,
"summary": "The submission introduces DriveArena, a high-fidelity closed-loop simulation platform designed for testing and developing autonomous driving agents in real-world ... | {
"rating": "3;6;6;6",
"rating_avg": 5.25,
"confidence": "4;4;3;5",
"confidence_avg": 4,
"soundness": "2;3;3;3",
"soundness_avg": 2.75,
"contribution": "2;3;3;3",
"contribution_avg": 2.75,
"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": "2026-01-14T22:16:00.809176"
} | {
"id": "kLZsCyg2HE",
"metareview": "This work proposes a closed-loop simulator for autonomous driving. The simulation involves two components: a traffic simulation system that generates traffic flow simulations and a multi-view image generation system for creating images based on the generated traffic and text pro... | {
"decision": "Reject"
} |
4T33izzFpK | 2407.12844v1 | metabench - A Sparse Benchmark of Reasoning and Knowledge in Large Language Models | {
"content": "## Abstract\n\nAbstract Large Language Models (LLMs) vary in their abilities on a range of tasks. Initiatives such as the Open LLM Leaderboard aim to quantify these differences with several large benchmarks (sets of test items to which an LLM can respond either correctly or incorrectly).\nHowever, high ... | [
{
"id": "ofgTEwn8MA",
"initial_rating": 6,
"confidence": 3,
"soundness": 4,
"contribution": 3,
"presentation": 4,
"summary": "This paper considers the six LLM benchmarks included in the Open LLM Leaderboard (ARC, GSM8K, HellaSwag, MMLU, TruthfulQA, and WinoGrande) and seeks to create a m... | {
"rating": "5;6;6;6",
"rating_avg": 5.75,
"confidence": "2;3;2;3",
"confidence_avg": 2.5,
"soundness": "2;4;3;4",
"soundness_avg": 3.25,
"contribution": "2;4;3;3",
"contribution_avg": 3,
"presentation": "2;3;3;4",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:00.809851"
} | {
"id": "tvOUxFk47w",
"metareview": "The authors were able to address all issues raised by the reviewers. All reviewers except one were positive about the work. That reviewer was not able to respond, but the authors addressed the issues by them by adding more experiments to evaluate the potential for data contamina... | {
"decision": "Accept (Poster)"
} |
4W1wTg7q9o | 2407.11965v2 | UrbanWorld: An Urban World Model for 3D City Generation | {
"content": "## Abstract\n\nAbstract Cities, as the essential environment of human life, encompass diverse physical elements such as buildings, roads and vegetation, which continuously interact with dynamic entities like people and vehicles.\nCrafting realistic, interactive 3D urban environments is essential for nur... | [
{
"id": "jn3hXcIyN4",
"initial_rating": 5,
"confidence": 5,
"soundness": 2,
"contribution": 2,
"presentation": 3,
"summary": "This paper proposes a generative urban world model that can automatically create a customized, realistic, and interactive 3D urban world with flexible control con... | {
"rating": "3;5;5;5",
"rating_avg": 4.5,
"confidence": "5;5;3;5",
"confidence_avg": 4.5,
"soundness": "2;3;3;2",
"soundness_avg": 2.5,
"contribution": "1;2;2;2",
"contribution_avg": 1.75,
"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": "2026-01-14T22:16:00.810751"
} | {
"id": "6eKBKhBKZZ",
"metareview": "**Summary**\n\nThe paper presents UrbanWorld, a generative world model for creating interactive 3D urban worlds. The generation consists of four stages: 1) generation of untextured 3D layouts based on input 2D layout, 2) using MLLM to generate textual descriptions detailing the ... | {
"decision": "Reject"
} |
4X9RpKH4Ls | 2408.14915v1 | Can Transformers Do Enumerative Geometry? | {
"content": "## Abstract\n\nAbstract. How can Transformers model and learn enumerative geometry? What is a robust procedure for using Transformers in abductive knowledge discovery within a mathematician-machine collaboration?\nIn this work, we introduce a new paradigm in computational enumerative geometry in analyzi... | [
{
"id": "exo9C2nJkK",
"initial_rating": 5,
"confidence": 3,
"soundness": 4,
"contribution": 3,
"presentation": 3,
"summary": "The paper proposes and tests the usage of transformers in the field of enumerative geometry, specifically regarding topological recursions and $\\psi$-class inter... | {
"rating": "3;3;5;5",
"rating_avg": 4,
"confidence": "2;4;2;3",
"confidence_avg": 2.75,
"soundness": "2;3;3;2",
"soundness_avg": 2.5,
"contribution": "1;2;3;3",
"contribution_avg": 2.25,
"presentation": "2;4;2;2",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:00.811401"
} | {
"id": "KJBmbuXu12",
"metareview": "**Summary:** \nThis paper introduces a Transformer-based model, DynamicFormer, with a novel Dynamic Range Activator (DRA) activation function tailored to model recursive functions with high variance and factorial growth. The study applies this methodology to computational enume... | {
"decision": "Accept (Poster)"
} |
4ZX2a3OKEV | 2411.05228v2 | Solving hidden monotone variational inequalities with surrogate losses | {
"content": "## Abstract\n\nAbstract Deep learning has proven to be effective in a wide variety of loss minimization problems.\nHowever, many applications of interest, like minimizing projected Bellman error and min-max optimization, cannot be modelled as minimizing a scalar loss function but instead correspond to s... | [
{
"id": "XZyKLauZsI",
"initial_rating": 8,
"confidence": 3,
"soundness": 4,
"contribution": 3,
"presentation": 3,
"summary": "This paper introduces a new method for optimizing variational inequalities with hidden structure by optimizing a series of surrogate losses, thereby extending pre... | {
"rating": "3;6;8;8",
"rating_avg": 6.25,
"confidence": "3;3;4;3",
"confidence_avg": 3.25,
"soundness": "2;3;4;4",
"soundness_avg": 3.25,
"contribution": "2;3;4;3",
"contribution_avg": 3,
"presentation": "2;2;3;3",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:00.812363"
} | {
"id": "h5UW9Dhp5d",
"metareview": "This paper studies a class of variational inequalities by assuming hidden monotonicity. The paper proposes a surrogate-based approach by constructing a surrogate and employ any optimizer to ensure a sufficient decrease condition. The paper provides some convergence analysis unde... | {
"decision": "Accept (Poster)"
} |
4a9doRh3Jv | 2406.12295v2 | Fast and Slow Generating: An Empirical Study on Large and Small Language Models Collaborative Decoding | {
"content": "## Abstract\n\nAbstract Large Language Models (LLMs) exhibit impressive capabilities across various applications but encounter substantial challenges such as high inference latency, considerable training costs, and the generation of hallucinations. Collaborative decoding between large and small language... | [
{
"id": "jtUyzkDKm0",
"initial_rating": 8,
"confidence": 4,
"soundness": 3,
"contribution": 4,
"presentation": 3,
"summary": "The paper explores collaborative decoding strategies between large language models (LLMs) and small language models (SLMs). The authors introduce the FS-GEN frame... | {
"rating": "3;6;6;8",
"rating_avg": 5.75,
"confidence": "3;4;3;4",
"confidence_avg": 3.5,
"soundness": "1;3;2;3",
"soundness_avg": 2.25,
"contribution": "2;2;3;4",
"contribution_avg": 2.75,
"presentation": "1;2;3;3",
"presentation_avg": 2.25
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:00.813339"
} | {
"id": "LUDv3mAYjK",
"metareview": "The paper examines collaborative decoding, where small and large language models work together during the decoding process. It unifies three techniques: proxy tuning, speculative decoding, and contrastive decoding, framing them through the FS-GEN (Fast and Slow Generating) frame... | {
"decision": "Reject"
} |
4anfpHj0wf | 2410.22493v1 | Unlocking Point Processes through Point Set Diffusion | {
"content": "## Abstract\n\nAbstract Point processes model the distribution of random point sets in mathematical spaces, such as spatial and temporal domains, with applications in fields like seismology, neuroscience, and economics.\nExisting statistical and machine learning models for point processes are predominan... | [
{
"id": "HLqXVnjxZa",
"initial_rating": 5,
"confidence": 5,
"soundness": 3,
"contribution": 2,
"presentation": 4,
"summary": "This paper proposes a novel modeling approach to point processes via diffusion on point sets (discrete events), addressing the reliance on the intensity function ... | {
"rating": "5;6;6;8",
"rating_avg": 6.25,
"confidence": "5;3;4;3",
"confidence_avg": 3.75,
"soundness": "3;3;3;3",
"soundness_avg": 3,
"contribution": "2;3;3;3",
"contribution_avg": 2.75,
"presentation": "4;3;3;4",
"presentation_avg": 3.5
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:00.813977"
} | {
"id": "TlVnAABhuY",
"metareview": "This paper proposes an interesting diffusion-based method for modeling and sampling point sets of point process distributions. For spatio-temporal point processes (STPP) the dominant approach is to apply autoregressive models that move points across time. This work compares to t... | {
"decision": "Accept (Poster)"
} |
4cQVUNpPkt | 2407.01494v1 | FOLEYCRAFTER: BRING SILENT VIDEOS TO LIFE WITH LIFELIKE AND SYNCHRONIZED SOUNDS | {
"content": "## Abstract\n\nAbstract We study Neural Foley, the automatic generation of high-quality sound effects synchronizing with videos, enabling an immersive audio-visual experience.\nDespite its wide range of applications, existing approaches encounter limitations when it comes to simultaneously synthesizing ... | [
{
"id": "JR5VnoJWwV",
"initial_rating": 8,
"confidence": 4,
"soundness": 2,
"contribution": 3,
"presentation": 4,
"summary": "This paper presents a new video-to-audio model, featured by the semantic adapter and temporal adapter. The proposed model uses the [Auffusion](https://arxiv.org/a... | {
"rating": "3;6;6",
"rating_avg": 5,
"confidence": "4;1;4",
"confidence_avg": 3,
"soundness": "2;3;2",
"soundness_avg": 2.3333333333333335,
"contribution": "2;3;2",
"contribution_avg": 2.3333333333333335,
"presentation": "1;3;3",
"presentation_avg": 2.3333333333333335
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:00.814651"
} | {
"id": "IvbQ5BQoUi",
"metareview": "The paper presents a framework: FoleyCrafter for adding foley sound effects to videos. The key innovations are: i) to enhance an audio latent diffusion model that is conditioned on text to also include video semantics via incorporating video features using cross-attention layers... | {
"decision": "Reject"
} |
4dAgG8ma3B | 2410.06502v1 | Chemistry-Inspired Diffusion with Non-Differentiable Guidance | {
"content": "## Abstract\n\nAbstract Recent advances in diffusion models have shown remarkable potential in the conditional generation of novel molecules. These models can be guided in two ways: (i) explicitly, through additional features representing the condition, or (ii) implicitly, using a property predictor. Ho... | [
{
"id": "rqdr3Zf6wq",
"initial_rating": 5,
"confidence": 4,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "The paper proposes CHEMGUIDE, a approach that uses non-differentiable guidance for the conditional generation of molecular structures in diffusion models. CHEMGUI... | {
"rating": "3;5;5;8",
"rating_avg": 5.25,
"confidence": "5;3;4;4",
"confidence_avg": 4,
"soundness": "2;2;3;4",
"soundness_avg": 2.75,
"contribution": "1;2;2;3",
"contribution_avg": 2,
"presentation": "2;2;3;4",
"presentation_avg": 2.75
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:00.815330"
} | {
"id": "4umnGLHieI",
"metareview": "The paper proposes a novel guidance technique for molecule generation. Namely, it targets generating molecules closer to their equilibrium state (with lower force magnitude) by adding the proposed guidance term to the diffusion model. The main challenge of this problem is that t... | {
"decision": "Accept (Poster)"
} |
4dAhjhm2Mm | 2408.16765v1 | A Score-Based Density Formula, with Applications in Diffusion Generative Models | {
"content": "## Abstract\n\nAbstract Score-based generative models (SGMs) have revolutionized the field\nof generative modeling, achieving unprecedented success in generating\nrealistic and diverse content. Despite empirical advances, the theoretical basis for why optimizing the evidence lower bound (ELBO) on the lo... | [
{
"id": "rse7rIDkLs",
"initial_rating": 3,
"confidence": 4,
"soundness": 2,
"contribution": 1,
"presentation": 2,
"summary": "This paper considers a density formula based on score estimation to analyze Denoising Diffusion Probabilistic Models (DDPMs). Using this formula, the authors prov... | {
"rating": "3;3;3;3",
"rating_avg": 3,
"confidence": "4;3;3;4",
"confidence_avg": 3.5,
"soundness": "2;3;4;2",
"soundness_avg": 2.75,
"contribution": "2;2;2;1",
"contribution_avg": 1.75,
"presentation": "1;2;2;2",
"presentation_avg": 1.75
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:00.816279"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
4es2oO9tw1 | 2410.16208v2 | Compute-Constrained Data Selection | {
"content": "## Abstract\n\nAbstract Data selection can reduce the amount of training data needed to finetune LLMs; however, the efficacy of data selection scales directly with its compute. Motivated by the practical challenge of compute-constrained finetuning, we consider the setting in which both the cost of selec... | [
{
"id": "Dm65QZWfdq",
"initial_rating": 8,
"confidence": 4,
"soundness": 4,
"contribution": 4,
"presentation": 4,
"summary": "Surveys and experimentally compares different data selection methods for LLM fine-tuning, and reasonably and quantitatively concludes that only rather cheap metho... | {
"rating": "3;5;6;8",
"rating_avg": 5.5,
"confidence": "3;3;3;4",
"confidence_avg": 3.25,
"soundness": "1;3;2;4",
"soundness_avg": 2.5,
"contribution": "1;3;1;4",
"contribution_avg": 2.25,
"presentation": "2;3;3;4",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:00.817185"
} | {
"id": "zhIkBO50gX",
"metareview": "The paper studies the cost of data selection by formalising the trade-off between data selection cost and training gain. The authors validate their insights experimentally. They show for instance that one has to pay attention to the computational cost of sophisticated model sele... | {
"decision": "Accept (Poster)"
} |
4fyg68nmd7 | 2411.05712v1 | Scaling Laws for Task-Optimized Models of the Primate Visual Ventral Stream | {
"content": "## Abstract\n\nAbstract When trained on large-scale object classification datasets, certain artificial neural network models begin to approximate core object recognition (COR) behaviors and neural response patterns in the primate visual ventral stream (VVS). While recent machine learning advances sugges... | [
{
"id": "2hc94wNRmF",
"initial_rating": 3,
"confidence": 4,
"soundness": 2,
"contribution": 2,
"presentation": 4,
"summary": "The authors investigate so-called neural scaling laws for predicting visual behavior and neural activity. \"Scaling laws\" are empirical trends that show a relati... | {
"rating": "3;5;6;6",
"rating_avg": 5,
"confidence": "4;4;3;4",
"confidence_avg": 3.75,
"soundness": "2;3;3;3",
"soundness_avg": 2.75,
"contribution": "2;2;2;2",
"contribution_avg": 2,
"presentation": "4;4;3;3",
"presentation_avg": 3.5
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:00.817918"
} | {
"id": "6kdTe6zemY",
"metareview": "The paper dives deep into the alignment of neural network models and neural response patterns in the visual ventral system. After checking myself, the quality of the paper and the visuals is top, and the findings are indeed intriguing and thought provoking, showing a great deal ... | {
"decision": "Reject"
} |
4hFT4rfG40 | 2410.02143v1 | Plug-and-Play Controllable Generation for Discrete Masked Models | {
"content": "## Abstract\n\nAbstract This article makes discrete masked models for the generative modeling of discrete data controllable. The goal is to generate samples of a discrete random variable that adheres to a posterior distribution, satisfies specific constraints, or optimizes a reward function. This method... | [
{
"id": "SqjCr8cRke",
"initial_rating": 1,
"confidence": 4,
"soundness": 1,
"contribution": 1,
"presentation": 1,
"summary": "This paper tackles the problem of performing conditional generation of discrete structures via masked generative models. They propose a general-purpose approach f... | {
"rating": "1;3;5;6",
"rating_avg": 3.75,
"confidence": "4;3;4;3",
"confidence_avg": 3.5,
"soundness": "1;2;3;4",
"soundness_avg": 2.5,
"contribution": "1;1;2;2",
"contribution_avg": 1.5,
"presentation": "1;3;4;3",
"presentation_avg": 2.75
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:00.818527"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
4hPwLg7zD3 | 2410.22269v1 | Fourier Head: Helping Large Language Models Learn Complex Probability Distributions | {
"content": "## Abstract\n\nAbstract As the quality of large language models has improved, there has been increased interest in using them to model non-linguistic tokens.\nFor example, the Decision Transformer recasts agentic decision making as a sequence modeling problem, using a decoder-only LLM to model the distr... | [
{
"id": "BonP0FB7Oo",
"initial_rating": 5,
"confidence": 3,
"soundness": 2,
"contribution": 2,
"presentation": 2,
"summary": "The paper proposes a Fourier Head based on the Fourier series as a replacement for the usual linear classification head to induce continuous densities across the ... | {
"rating": "5;5;5;6",
"rating_avg": 5.25,
"confidence": "4;4;3;3",
"confidence_avg": 3.5,
"soundness": "3;2;2;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": "2026-01-14T22:16:00.819684"
} | {
"id": "Nu2NqhjjRT",
"metareview": "This paper proposes a new mechanism to model discrete distributions using Fourier representations. The motivation is to better model the underlying low-frequency signals in the Fourier space. Empirical validation is done in two domains for offline RL and time series foundation m... | {
"decision": "Accept (Poster)"
} |
4hdDPa9bpI | 2410.04655v2 | Graph Fourier Neural Kernels (G-FuNK): Learning Solutions of Nonlinear Diffusive Parametric PDEs on Multiple Domains | {
"content": "## Abstract\n\nAbstract Predicting the time-dependent dynamics of complex systems governed by non-linear partial differential equations (PDEs), with varying parameters and domains, is a difficult problem that is motivated by applications in many fields. We introduce a novel family of neural operators ba... | [
{
"id": "fqEBagVuJ0",
"initial_rating": 3,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "The authors proposes an interesting surrogate model that combines Graph as the discretization method for Fourier neural operators. I believe the manuscript may be... | {
"rating": "3;3;5;8",
"rating_avg": 4.75,
"confidence": "3;3;3;4",
"confidence_avg": 3.25,
"soundness": "2;3;2;3",
"soundness_avg": 2.5,
"contribution": "2;3;2;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": "2026-01-14T22:16:00.820512"
} | {
"id": "U6YWcgF7t5",
"metareview": "The paper introduces a novel family of neural operators, termed Graph Fourier Neural Kernels (G-FuNK), designed to learn the temporal dynamics of diffusive PDEs across multiple anisotropic domains with varying parameters. The method embeds geometric and directional information a... | {
"decision": "Reject"
} |
4hp2bVdaHU | 2410.10984v1 | Data-Aware Training Quality Monitoring and Certification for Reliable Deep Learning | {
"content": "## Abstract\n\nAbstract Deep learning models excel at capturing complex representations through sequential layers of linear and non-linear transformations, yet their inherent black-box nature and multi-modal training landscape raise critical concerns about reliability, robustness, and safety, particular... | [
{
"id": "w4Fk4aoM9J",
"initial_rating": 3,
"confidence": 2,
"soundness": 1,
"contribution": 2,
"presentation": 2,
"summary": "This paper introduces YES training bounds, a framework for real-time, data-aware certification and monitoring of neural network training. The framework evaluates ... | {
"rating": "3;3;3;5",
"rating_avg": 3.5,
"confidence": "3;4;2;2",
"confidence_avg": 2.75,
"soundness": "2;1;1;2",
"soundness_avg": 1.5,
"contribution": "2;1;2;2",
"contribution_avg": 1.75,
"presentation": "2;2;2;2",
"presentation_avg": 2
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:00.821284"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
4ihkxIeTFH | 2405.12807v11 | FAdam: Adam is a natural gradient optimizer using diagonal empirical Fisher information | {
"content": "## Abstract\n\nAbstract This paper establishes a mathematical foundation for the Adam optimizer, elucidating its connection to natural gradient descent through Riemannian and information geometry.\nWe provide an accessible and detailed analysis of the diagonal empirical Fisher information matrix (FIM) i... | [
{
"id": "bszfdJbm9M",
"initial_rating": 3,
"confidence": 5,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "This paper studies the connection between Adam optimizer and natural gradient descent by leveraging techniques from Riemannian geometry. Based on this, the author... | {
"rating": "3;3;3;3;3",
"rating_avg": 3,
"confidence": "3;2;4;4;5",
"confidence_avg": 3.6,
"soundness": "2;2;1;2;3",
"soundness_avg": 2,
"contribution": "1;2;2;2;2",
"contribution_avg": 1.8,
"presentation": "1;3;2;2;3",
"presentation_avg": 2.2
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:00.822210"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
4ikjWBs3tE | 2403.06925v1 | Transformers Learn Low Sensitivity Functions: Investigations and Implications | {
"content": "## Abstract\n\nAbstract Transformers achieve state-of-the-art accuracy and robustness across many tasks, but an understanding of the inductive biases that they have and how those biases are different from other neural network architectures remains elusive. Various neural network architectures such as fu... | [
{
"id": "4RET9VLNgp",
"initial_rating": 8,
"confidence": 2,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "The paper considers implicit biases for the transformer architecture. They describe sensitivity of a function as the change in the function value averaged over al... | {
"rating": "3;5;8;8",
"rating_avg": 6,
"confidence": "4;3;3;2",
"confidence_avg": 3,
"soundness": "2;1;4;3",
"soundness_avg": 2.5,
"contribution": "2;3;3;3",
"contribution_avg": 2.75,
"presentation": "3;3;4;3",
"presentation_avg": 3.25
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:00.823105"
} | {
"id": "POxRwjDTKI",
"metareview": "This paper extends the original notion of sensitivity in Boolean function analysis to an empirically measurable quantity and study it on transformers. The main conclusions contain three points, including lower sensitivity correlates well with robustness (this is almost by defini... | {
"decision": "Accept (Poster)"
} |
4muXQ5r8Ol | 2406.03070v1 | A-Bench: Are LMMs Masters at Evaluating AI-generated Images? | {
"content": "## Abstract\n\nAbstract How to accurately and efficiently assess AI-generated images (AIGIs) remains a critical challenge for generative models. Given the high costs and extensive time commitments required for user studies, many researchers have turned towards employing large multi-modal models (LMMs) a... | [
{
"id": "qly712uEDo",
"initial_rating": 3,
"confidence": 5,
"soundness": 2,
"contribution": 1,
"presentation": 2,
"summary": "Due to the existing evaluation models' inability to effectively assess the performance of AIGI tasks, more and more researchers are turning to LMMs for evaluating... | {
"rating": "3;5;6;8",
"rating_avg": 5.5,
"confidence": "5;3;4;5",
"confidence_avg": 4.25,
"soundness": "2;2;3;4",
"soundness_avg": 2.75,
"contribution": "1;2;3;3",
"contribution_avg": 2.25,
"presentation": "2;2;4;4",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:00.823998"
} | {
"id": "ZMIYFroFzM",
"metareview": "Summary\nThe paper examines whether multimodal models can evaluate image generation models. The authors propose a benchmark, A-Bench, that can evaluate both the text alignment and the image quality of generation models. A-Bench is a diagnostic benchmark, and the authors use diff... | {
"decision": "Accept (Poster)"
} |
4sDicVEy6M | 2409.02335v1 | What Do You See in Common? Learning Hierarchical Prototypes over Tree-of-Life to Discover Evolutionary Traits | {
"content": "## Abstract\n\nAbstract A grand challenge in biology is to discover evolutionary traits—features of organisms common to a group of species with a shared ancestor in the tree of life (also referred to as phylogenetic tree). With the growing availability of image repositories in biology, there is a tremen... | [
{
"id": "C5rsG4RdmB",
"initial_rating": 6,
"confidence": 4,
"soundness": 2,
"contribution": 3,
"presentation": 3,
"summary": "This paper applies deep learning techniques to uncover evolutionary traits in biological data. It leverages contrastive and orthogonality losses to facilitate hie... | {
"rating": "5;6;6;6;6",
"rating_avg": 5.8,
"confidence": "3;2;3;3;4",
"confidence_avg": 3,
"soundness": "2;3;3;3;2",
"soundness_avg": 2.6,
"contribution": "3;2;3;2;3",
"contribution_avg": 2.6,
"presentation": "3;3;3;2;3",
"presentation_avg": 2.8
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:00.825215"
} | {
"id": "vzxgzCYqc8",
"metareview": "HComP-Net (Hierarchy aligned Commonality through Prototypical Networks) is a new machine learning framework that discovers evolutionary traits in species by analysing images and learning hierarchical prototypes aligned with phylogenetic trees. The system's key contribution is it... | {
"decision": "Accept (Poster)"
} |
4tiTQ33sDH | 2305.03977v3 | Unlocking the Power of GANs in Non-Autoregressive Text Generation | {
"content": "## Abstract\n\nAbstract Generative Adversarial Networks (GANs) have been studied in text generation to tackle the exposure bias problem. Despite their remarkable development, they adopt autoregressive structures so suffering from high latency in both training and inference stages. Although GANs have pot... | [
{
"id": "qJ1gswmTm5",
"initial_rating": 3,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "The paper introduces a novel model called Adversarial Non-autoregressive Transformer (ANT) aimed at enhancing the efficiency and performance of Generative Adversa... | {
"rating": "3;3;3;5",
"rating_avg": 3.5,
"confidence": "2;5;3;4",
"confidence_avg": 3.5,
"soundness": "2;2;3;2",
"soundness_avg": 2.25,
"contribution": "2;2;3;3",
"contribution_avg": 2.5,
"presentation": "2;2;3;3",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:00.825860"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
4v4nmYWzBa | 2410.06665v1 | REVISITING MULTI-PERMUTATION EQUIVARIANCE THROUGH THE LENS OF IRREDUCIBLE REPRESENTATIONS | {
"content": "## Abstract\n\nAbstract This paper explores the characterization of equivariant linear layers for representations of permutations and related groups.\nUnlike traditional approaches, which address these problems using parameter-sharing, we consider an alternative methodology based on irreducible represen... | [
{
"id": "N2ms9vYJvU",
"initial_rating": 3,
"confidence": 3,
"soundness": 2,
"contribution": 2,
"presentation": 2,
"summary": "The paper introduces an alternative approach for characterizing equivariant linear layers in neural networks that process permutation and related group representa... | {
"rating": "3;6;6;6",
"rating_avg": 5.25,
"confidence": "3;4;3;4",
"confidence_avg": 3.5,
"soundness": "2;3;4;3",
"soundness_avg": 3,
"contribution": "2;2;2;3",
"contribution_avg": 2.25,
"presentation": "2;4;3;3",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:00.826586"
} | {
"id": "Oc1CyGKqGm",
"metareview": "This paper derives the existing equivariant models of Deep Sets, 2-IGNs, and Deep Weight Space networks, in terms of irreducible representations and Schur's lemma. The concept of the paper is interesting, and the theoretical contribution delivers a mathematical approach to under... | {
"decision": "Accept (Poster)"
} |
4vzGQcVUG8 | 2410.04638v1 | Provable weak-to-strong generalization via benign overfitting | {
"content": "## Abstract\n\nAbstract The classic teacher-student model in machine learning posits that a strong teacher supervises a weak student to improve the student’s capabilities.\nWe instead consider the inverted situation, where a weak teacher supervises a strong student with imperfect pseudolabels.\nThis par... | [
{
"id": "yMmRjPYOxZ",
"initial_rating": 6,
"confidence": 3,
"soundness": 4,
"contribution": 4,
"presentation": 3,
"summary": "The paper investigates weak-to-strong generalization in the setting of an overparameterized spiked covariance model with Gaussian covariates. The paper identifies... | {
"rating": "5;6;6;8",
"rating_avg": 6.25,
"confidence": "2;3;3;2",
"confidence_avg": 2.5,
"soundness": "3;3;4;3",
"soundness_avg": 3.25,
"contribution": "2;3;4;3",
"contribution_avg": 3,
"presentation": "2;2;3;2",
"presentation_avg": 2.25
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:00.827843"
} | {
"id": "aQqkyAi7AH",
"metareview": "Motivated by recent work by Burns et al., this paper identifies a stylized setting where one can show weak to strong generalization theoretically. The weak model is trained on 'weak features' and this model is used to generate pseudolabels for more (unlabeled data). The strong m... | {
"decision": "Accept (Poster)"
} |
4w99NAikOE | 2410.07171v1 | IterComp: Iterative Composition-Aware Feedback Learning from Model Gallery for Text-to-Image Generation | {
"content": "## Abstract\n\nAbstract Advanced diffusion models like RPG, Stable Diffusion 3 and FLUX have made notable strides in compositional text-to-image generation. However, these methods typically exhibit distinct strengths for compositional generation, with some excelling in handling attribute binding and oth... | [
{
"id": "ENKeelceX2",
"initial_rating": 6,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "This paper proposes IterComp, an iterative composition-aware reward-controlled framework. It introduces a model gallery and constructs a high-quality composition-... | {
"rating": "5;6;6;6;6",
"rating_avg": 5.8,
"confidence": "3;4;4;4;3",
"confidence_avg": 3.6,
"soundness": "2;3;3;3;3",
"soundness_avg": 2.8,
"contribution": "2;3;2;3;3",
"contribution_avg": 2.6,
"presentation": "3;3;3;3;3",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:00.828931"
} | {
"id": "nwxUTihai1",
"metareview": "All reviewers agree to accept the paper. Reviewers appreciate the novel framework and significant performance improvement. Please be sure to address the reviewers' comments in the final version.",
"additional_comments": "All reviewers agree to accept the paper."
} | {
"decision": "Accept (Poster)"
} |
4wmf3Ffhl2 | 2405.13753v3 | A Dynamic Model of Performative Human-ML Collaboration: Theory and Empirical Evidence | {
"content": "## Abstract\n\nAbstract Machine learning (ML) models are increasingly used in various applications, from recommendation systems in e-commerce to diagnosis prediction in healthcare.\nIn this paper, we present a novel dynamic framework for thinking about the deployment of ML models in a performative, huma... | [
{
"id": "AzayiHOXsf",
"initial_rating": 3,
"confidence": 3,
"soundness": 3,
"contribution": 1,
"presentation": 2,
"summary": "* This paper examines Human-ML collaboration under performative prediction settings through theoretical analysis and an empirical experiment on ML-assisted combin... | {
"rating": "3;5;5;5",
"rating_avg": 4.5,
"confidence": "3;3;2;3",
"confidence_avg": 2.75,
"soundness": "3;3;2;3",
"soundness_avg": 2.75,
"contribution": "1;2;2;3",
"contribution_avg": 2,
"presentation": "2;3;2;2",
"presentation_avg": 2.25
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:00.829618"
} | {
"id": "xWXx55ayzB",
"metareview": "The paper presents a dynamic framework for performative human-ML collaboration, modeling how human decisions influenced by ML predictions can alter the data-generating process. The authors propose a utility-based theoretical model using collaborative characteristic functions to ... | {
"decision": "Reject"
} |
4wtcXV0kbi | 2410.03464v1 | S7: Selective and Simplified State Space Layers for Sequence Modeling | {
"content": "## Abstract\n\nAbstract A central challenge in sequence modeling is efficiently handling tasks with extended contexts. While recent state-space models (SSMs) have made significant progress in this area, they often lack input-dependent filtering or require substantial increases in model complexity to han... | [
{
"id": "QzawM73X5C",
"initial_rating": 5,
"confidence": 4,
"soundness": 3,
"contribution": 2,
"presentation": 2,
"summary": "This paper proposed a new SSM called S7 for sequence modeling. It combines the strengths of S5 (simpler structure) and S7 (input-dependent state transitions) and ... | {
"rating": "3;3;3;5",
"rating_avg": 3.5,
"confidence": "5;5;4;4",
"confidence_avg": 4.5,
"soundness": "2;2;2;3",
"soundness_avg": 2.25,
"contribution": "1;2;1;2",
"contribution_avg": 1.5,
"presentation": "2;1;2;2",
"presentation_avg": 1.75
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:00.830414"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
4xWQS2z77v | 2411.07729v1 | Exploring The Loss Landscape Of Regularized Neural Networks Via Convex Duality | {
"content": "## Abstract\n\nAbstract We discuss several aspects of the loss landscape of regularized neural networks: the structure of stationary points, connectivity of optimal solutions, path with nonincreasing loss to arbitrary global optimum, and the nonuniqueness of optimal solutions, by casting the problem int... | [
{
"id": "PrQin5TgUr",
"initial_rating": 8,
"confidence": 5,
"soundness": 3,
"contribution": 3,
"presentation": 4,
"summary": "The authors present a deep and novel analysis of the loss landscape and a solution in the context of regularized neural networks. They also show that the topology... | {
"rating": "5;6;6;6;8",
"rating_avg": 6.2,
"confidence": "3;4;2;2;5",
"confidence_avg": 3.2,
"soundness": "3;3;3;3;3",
"soundness_avg": 3,
"contribution": "3;3;3;3;3",
"contribution_avg": 3,
"presentation": "1;2;2;2;4",
"presentation_avg": 2.2
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Oral",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:00.831524"
} | {
"id": "fewshNicIW",
"metareview": "This paper studies the loss landscape of regularized ReLU networks, focusing on the structure of stationary points, the connectivity of optimal solutions and the non uniqueness of optimal solutions. The authors start by considering a two-layer network with scalar output, and cha... | {
"decision": "Accept (Oral)"
} |
4xbwWerxvZ | 2403.12063v2 | Consistency Model is an Effective Posterior Sample Approximation for Diffusion Inverse Solvers | {
"content": "## Abstract\n\nAbstract Diffusion Inverse Solvers (DIS) are designed to sample from the conditional distribution p θ ( X 0 | y ) subscript 𝑝 𝜃 conditional subscript 𝑋 0 𝑦 p_{\\theta}(X_{0}|y) italic_p start_POSTSUBSCRIPT italic_θ end_POSTSUBSCRIPT ( italic_X start_POSTSUBSCRIPT 0 end_POSTSUBSCRIPT... | [
{
"id": "GG4PmbE9Yv",
"initial_rating": 5,
"confidence": 3,
"soundness": 2,
"contribution": 2,
"presentation": 2,
"summary": "The paper proposes using consistency models (CMs) to solve inverse problems with diffusion models. In diffusion inverse problems, common approaches need to go fro... | {
"rating": "5;5;5;6",
"rating_avg": 5.25,
"confidence": "4;3;3;5",
"confidence_avg": 3.75,
"soundness": "3;3;2;3",
"soundness_avg": 2.75,
"contribution": "3;3;2;2",
"contribution_avg": 2.5,
"presentation": "3;3;2;2",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:00.832715"
} | {
"id": "uFHUxpJOm6",
"metareview": "The paper explores the use of consistency models (CMs) to address inverse problems in diffusion models by replacing the traditional expectation computation step with a CM-based approach. This allows for faster sampling from the probability flow ODE (PF-ODE) while avoiding costly... | {
"decision": "Reject"
} |
4y4t7yOvJO | 2409.20447v1 | POMONAG: Pareto-Optimal Many-Objective Neural Architecture Generator | {
"content": "## Abstract\n\nAbstract Neural Architecture Search (NAS) automates the design of neural network architectures, minimising dependence on human expertise and iterative experimentation. While NAS methods are often computationally intensive and dataset-specific, employing auxiliary predictors to estimate ar... | [
{
"id": "eMLqdgmxUP",
"initial_rating": 3,
"confidence": 5,
"soundness": 2,
"contribution": 2,
"presentation": 2,
"summary": "This paper introduces POMONAG, an extension to DiffusionNAG that applies a many-objective diffusion model to optimize neural architecture generation for many-obje... | {
"rating": "1;3;5;5;6",
"rating_avg": 4,
"confidence": "5;5;5;3;3",
"confidence_avg": 4.2,
"soundness": "2;2;3;3;3",
"soundness_avg": 2.6,
"contribution": "2;2;2;2;3",
"contribution_avg": 2.2,
"presentation": "1;2;2;3;4",
"presentation_avg": 2.4
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:00.833549"
} | {
"id": "ghrn26tUTy",
"metareview": "This study proposes POMONAG, extending DiffusionNAG for multi-objective optimization. While the method is clearly presented with experiments across 15 datasets, reviewers noted limited novelty, as the approach largely builds on DiffusionNAG with minor extensions. The lack of com... | {
"decision": "Reject"
} |
4ytRL3HJrq | 2311.13721v5 | Nova: Generative Language Models for Assembly Code with Hierarchical Attention and Contrastive Learning | {
"content": "## Abstract\n\nAbstract Binary code analysis is the foundation of crucial tasks in the security domain; thus building effective binary analysis techniques is more important than ever. Large language models (LLMs) although have brought impressive improvement to source code tasks, do not directly generali... | [
{
"id": "88KucOADA5",
"initial_rating": 5,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 2,
"summary": "The paper presents a way of training an LLM to improve its performance on tasks that require understanding of assembly code, in particular code decompilation, and... | {
"rating": "3;5;5;6;8",
"rating_avg": 5.4,
"confidence": "4;3;4;3;3",
"confidence_avg": 3.4,
"soundness": "3;3;3;2;3",
"soundness_avg": 2.8,
"contribution": "2;2;3;3;3",
"contribution_avg": 2.6,
"presentation": "3;3;2;2;4",
"presentation_avg": 2.8
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:00.834417"
} | {
"id": "Xvle72TClz",
"metareview": "This paper studies using LLMs for binary code analysis. Despite many works used LLMs on source code tasks, they cannot be directly generalized to assembly code. The authors proposed an attention mechanism to capture the semantics in binary code for LLM training. Their method can... | {
"decision": "Accept (Poster)"
} |
4z3IguA4Zg | 2410.11779v1 | MLLM can see? Dynamic Correction Decoding for Hallucination Mitigation | {
"content": "## Abstract\n\nAbstract Multimodal Large Language Models (MLLMs) frequently exhibit hallucination phenomena, but the underlying reasons remain poorly understood. In this paper, we present an empirical analysis and find that, although MLLMs incorrectly generate the objects in the final output, they are a... | [
{
"id": "REb8ztCB1E",
"initial_rating": 6,
"confidence": 5,
"soundness": 3,
"contribution": 3,
"presentation": 4,
"summary": "The paper introduces DeCo (Dynamic Correction Decoding), a decoding technique to mitigate hallucinations in Multimodal Large Language Models (MLLMs). The authors ... | {
"rating": "5;5;6;6",
"rating_avg": 5.5,
"confidence": "4;4;3;5",
"confidence_avg": 4,
"soundness": "3;2;3;3",
"soundness_avg": 2.75,
"contribution": "3;2;3;3",
"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": "2026-01-14T22:16:00.835190"
} | {
"id": "c7mGjg71uv",
"metareview": "This works shows that though MLLMs could generate incorrect outputs in the final layer, they could effectively recognise objects in preceding layers. Thus, this work introduces a dynamic correction decoding approach for MLLMs to adaptively choose relevant layers and integrate th... | {
"decision": "Accept (Poster)"
} |
4zygH3k8Zr | 2410.01239v1 | Replacement Learning: Training Vision Tasks with Fewer Learnable Parameters | {
"content": "## Abstract\n\nAbstract Traditional end-to-end deep learning models often enhance feature representation and overall performance by increasing the depth and complexity of the network during training. However, this approach inevitably introduces issues of parameter redundancy and resource inefficiency, e... | [
{
"id": "RoBbjcYigt",
"initial_rating": 8,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "This paper introduces Replacement Learning, a novel training technique aimed at reducing the number of learnable parameters in deep learning models while maintain... | {
"rating": "1;3;5;5;8",
"rating_avg": 4.4,
"confidence": "4;4;4;5;3",
"confidence_avg": 4,
"soundness": "1;2;2;2;3",
"soundness_avg": 2,
"contribution": "1;2;3;3;3",
"contribution_avg": 2.4,
"presentation": "1;3;2;3;3",
"presentation_avg": 2.4
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:00.836555"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
506BjJ1ziZ | 2410.10894v1 | COME: Test-time Adaption by Conservatively Minimizing Entropy | {
"content": "## Abstract\n\nAbstract Machine learning models must continuously self-adjust themselves for novel data distribution in the open world. As the predominant principle, entropy minimization (EM) has been proven to be a simple yet effective cornerstone in existing test-time adaption (TTA) methods. While unf... | [
{
"id": "Vp0vz2Rkif",
"initial_rating": 8,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 4,
"summary": "This paper addresses the model collapse of the popular Entropy Minimization algorithm for Test-Time Adaptation. Motivated by the observation that the amplificatio... | {
"rating": "3;5;5;8",
"rating_avg": 5.25,
"confidence": "3;3;4;4",
"confidence_avg": 3.5,
"soundness": "2;3;3;3",
"soundness_avg": 2.75,
"contribution": "2;2;2;3",
"contribution_avg": 2.25,
"presentation": "2;3;3;4",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:00.837243"
} | {
"id": "WmHKPNrApS",
"metareview": "This paper studies the issue of model collapse in entropy minimization algorithms for test-time adaptation. It proproses a entropy minimization approach that models prediction uncertainty by a Dirichlet prior distribution over model predictions. This method regularizes the model... | {
"decision": "Accept (Poster)"
} |
50RNY6uM2Q | 2406.17770v2 | MG-LLaVA: Towards Multi-Granularity Visual Instruction Tuning | {
"content": "## Abstract\n\nAbstract Multi-modal large language models (MLLMs) have made significant strides in various visual understanding tasks.\nHowever, the majority of these models are constrained\nto process low-resolution images,\nwhich limits their effectiveness in perception tasks that necessitate detailed... | [
{
"id": "BpSahESDjl",
"initial_rating": 5,
"confidence": 5,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "This paper presents an MLLM architecture to improve the multi-granularity visual understanding abilities of multimodal models. The method follows the idea propose... | {
"rating": "3;5;5;5;5",
"rating_avg": 4.6,
"confidence": "5;5;3;5;5",
"confidence_avg": 4.6,
"soundness": "2;2;3;3;3",
"soundness_avg": 2.6,
"contribution": "2;2;2;2;2",
"contribution_avg": 2,
"presentation": "3;2;4;3;3",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:00.837916"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
50cmx4SrkM | 2312.12676v2 | Bayesian Analysis of Combinatorial Gaussian Process Bandits | {
"content": "## Abstract\n\nAbstract We consider the combinatorial volatile Gaussian process (GP) semi-bandit problem. Each round, an agent is provided a set of available base arms and must select a subset of them to maximize the long-term cumulative reward. We study the Bayesian setting and provide novel Bayesian c... | [
{
"id": "nkjnwwAI9g",
"initial_rating": 5,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 2,
"summary": "The authors derive Bayesian regret bounds for various algorithms applied to combinatorial volatile GP semi-bandit problems. Specifically, the authors derive regre... | {
"rating": "5;5;5;8",
"rating_avg": 5.75,
"confidence": "2;4;3;3",
"confidence_avg": 3,
"soundness": "3;2;3;3",
"soundness_avg": 2.75,
"contribution": "2;1;3;3",
"contribution_avg": 2.25,
"presentation": "4;2;2;3",
"presentation_avg": 2.75
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:00.838703"
} | {
"id": "u885YE38XO",
"metareview": "This is a borderline paper with one reviewer being very positive and three reviewers somewhat critical. After having had a read through the reviews and discussions I believe that the criticisms are mostly about minor issues like the presentation of results. The paper presents us... | {
"decision": "Accept (Poster)"
} |
514rdneWOX | 2410.09962v2 | LongHalQA: Long-Context Hallucination Evaluation for MultiModal Large Language Models | {
"content": "## Abstract\n\nAbstract Hallucination, a phenomenon where multimodal large language models (MLLMs) tend to generate textual responses that are plausible but unaligned with the image, has become one major hurdle in various MLLM-related applications. Several benchmarks have been created to gauge the hallu... | [
{
"id": "fANbJtBgi9",
"initial_rating": 3,
"confidence": 4,
"soundness": 2,
"contribution": 2,
"presentation": 1,
"summary": "This paper proposes a new benchmark for evaluating hallucinations in multi-modal large language models (MLLMs).\nThe paper makes use of GPT4V to generate image-le... | {
"rating": "3;5;6;8",
"rating_avg": 5.5,
"confidence": "4;5;4;3",
"confidence_avg": 4,
"soundness": "2;2;2;3",
"soundness_avg": 2.25,
"contribution": "2;2;2;3",
"contribution_avg": 2.25,
"presentation": "1;2;3;3",
"presentation_avg": 2.25
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:00.839445"
} | {
"id": "rLbyeLxBdK",
"metareview": "The paper introduces LongHalQA, a new benchmark for evaluating hallucinations in multi-modal large language models (MLLMs). Hallucinations occur when a model generates text that misrepresents the input image. The benchmark tries to addresses the key limitations in existing evalu... | {
"decision": "Reject"
} |
51WraMid8K | 2410.03523v3 | A Probabilistic Perspective on Unlearning and Alignment for Large Language Models | {
"content": "## Abstract\n\nAbstract Comprehensive evaluation of Large Language Models (LLMs) is an open research problem. Existing evaluations rely on deterministic point estimates generated via greedy decoding. However, we find that deterministic evaluations fail to capture the whole output distribution of a model... | [
{
"id": "02CjSuqtuK",
"initial_rating": 10,
"confidence": 3,
"soundness": 4,
"contribution": 4,
"presentation": 3,
"summary": "In this paper authors address the problem of reliable unlearning in LLMs. First they introduce a problem, that evaluations based on deterministic point estimates... | {
"rating": "5;6;8;10",
"rating_avg": 7.25,
"confidence": "3;3;2;3",
"confidence_avg": 2.75,
"soundness": "3;3;4;4",
"soundness_avg": 3.5,
"contribution": "3;3;3;4",
"contribution_avg": 3.25,
"presentation": "2;2;3;3",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Oral",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:00.839957"
} | {
"id": "c0kNDLz1Dg",
"metareview": "This paper proposes a formal probabilistic evaluation framework for LLMs and designs new metrics with high-probability guarantees concerning the output distribution of a model. In addition, an unlearning loss based on entropy optimization and adaptive temperature scaling is also... | {
"decision": "Accept (Oral)"
} |
52XG8eexal | 2410.11687v1 | State-space models can learn in-context by gradient descent | {
"content": "## Abstract\n\nAbstract Deep state-space models (Deep SSMs) have shown capabilities for in-context learning on autoregressive tasks, similar to transformers.\nHowever, the architectural requirements and mechanisms enabling this in recurrent networks remain unclear.\nThis study demonstrates that state-sp... | [
{
"id": "rJdJx9V7fb",
"initial_rating": 3,
"confidence": 5,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "The paper investigates the ability of SSMs to perform ICL through gradient descent during the forward pass over recurrent steps. It provides a theoretical analysi... | {
"rating": "3;3;3;5",
"rating_avg": 3.5,
"confidence": "5;4;5;3",
"confidence_avg": 4.25,
"soundness": "3;4;3;2",
"soundness_avg": 3,
"contribution": "1;1;2;2",
"contribution_avg": 1.5,
"presentation": "2;3;3;3",
"presentation_avg": 2.75
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:00.840478"
} | {
"id": "bUsFIb22RB",
"metareview": "The paper proposes that ssms can perform in-context learning through gradient descent during their forward pass. The authors try to provide theoretical and empirical evidence that SSMs augmented with local self-attention can emulate gradient descent on implicit regression models... | {
"decision": "Reject"
} |
52x04chyQs | 2402.04836v2 | On the Completeness of Invariant Geometric Deep Learning Models | {
"content": "## Abstract\n\nAbstract Invariant models, one important class of geometric deep learning models, are capable of generating meaningful geometric representations by leveraging informative geometric features in point clouds. These models are characterized by their simplicity, good experimental results and ... | [
{
"id": "8pdrh5i12Z",
"initial_rating": 5,
"confidence": 3,
"soundness": 2,
"contribution": 4,
"presentation": 2,
"summary": "This paper explores the geometric completeness of a significant class of geometric deep learning models: invariant neural networks. These networks leverage invari... | {
"rating": "5;5;5;6",
"rating_avg": 5.25,
"confidence": "2;3;3;2",
"confidence_avg": 2.5,
"soundness": "2;3;2;3",
"soundness_avg": 2.5,
"contribution": "2;3;3;3",
"contribution_avg": 2.75,
"presentation": "3;1;2;2",
"presentation_avg": 2
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:00.841454"
} | {
"id": "V4VidEFjUD",
"metareview": "This paper analyzes the theoretical expressive power of invariant geometric deep learning models, focusing on their ability to represent point cloud geometry under fully-connected conditions. It establishes that DisGNN (message-passing neural networks incorporating distance) is ... | {
"decision": "Accept (Poster)"
} |
55oi1LCdDL | 2410.00911v1 | Dual Consolidation for Pre-Trained Model-Based Domain-Incremental Learning | {
"content": "## Abstract\n\nAbstract Domain-Incremental Learning (DIL) involves the progressive adaptation of a model to new concepts across different domains. While recent advances in pre-trained models provide a solid foundation for DIL, learning new concepts often results in the catastrophic forgetting of pre-tra... | [
{
"id": "NpajcA7ymV",
"initial_rating": 6,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "This paper introduces DUCT, a dual consolidation technique for domain-incremental learning (DIL) that effectively mitigates catastrophic forgetting. DUCT addresse... | {
"rating": "3;5;5;6;6",
"rating_avg": 5,
"confidence": "4;4;4;4;4",
"confidence_avg": 4,
"soundness": "2;3;2;3;3",
"soundness_avg": 2.6,
"contribution": "2;2;2;3;3",
"contribution_avg": 2.4,
"presentation": "3;2;2;3;3",
"presentation_avg": 2.6
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:00.842463"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
562B7aLi5X | 2407.01371v1 | Binary Losses for Density Ratio Estimation | {
"content": "## Abstract\n\nAbstract Estimating the ratio of two probability densities from finitely many observations of the densities, is a central problem in machine learning and statistics.\nA large class of methods constructs estimators from binary classifiers which distinguish observations from the two densiti... | [
{
"id": "YZ9ICVmgiU",
"initial_rating": 6,
"confidence": 2,
"soundness": 3,
"contribution": 2,
"presentation": 2,
"summary": "The authors characterize the set of loss functions that, when used in density ratio estimation for binary classification, lead to the minimization of a particular... | {
"rating": "3;3;6;6;8",
"rating_avg": 5.2,
"confidence": "2;2;3;3;2",
"confidence_avg": 2.4,
"soundness": "2;3;3;3;4",
"soundness_avg": 3,
"contribution": "1;1;3;3;3",
"contribution_avg": 2.2,
"presentation": "2;2;2;3;3",
"presentation_avg": 2.4
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:00.843359"
} | {
"id": "rQRerAwVlb",
"metareview": "The paper addresses an important problem, definitely in a proper way (no pun intended). In my decision to accept the paper, I have taken into account the initial opinions and their revised version, the responses of the authors, and the paper (I read it).\n\nIt is important to ac... | {
"decision": "Accept (Poster)"
} |
56mg1JFd3n | 2408.14906v1 | Writing in the Margins: Better Inference Patterns for Long-Context Retrieval | {
"content": "## 1 Introduction\n\nThe performance of Large Language Models (LLMs) tends to deteriorate when processing extensive inputs, a limitation linked directly to their fixed context window and attention mechanisms [li2023looglelongcontextlanguagemodels, liu2023lostmiddlelanguagemodels]. In particular, LLMs st... | [
{
"id": "CDaizZ1l4L",
"initial_rating": 10,
"confidence": 4,
"soundness": 3,
"contribution": 4,
"presentation": 4,
"summary": "The paper introduces a new inference methodology called \"writing in margins\" for long context tasks. The method builds upon the chunked prefill strategy (commo... | {
"rating": "3;5;6;10",
"rating_avg": 6,
"confidence": "2;4;2;4",
"confidence_avg": 3,
"soundness": "1;2;3;3",
"soundness_avg": 2.25,
"contribution": "1;2;3;4",
"contribution_avg": 2.5,
"presentation": "2;3;3;4",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:00.844146"
} | {
"id": "ampRSo47Ia",
"metareview": "This paper proposes a new method called Writing in the Margins (WiM) to improve inference for long-context retrieval. The strategy presented can be applied to different foundation models and requires minimal additional computation. Although a reviewer gave a very high score, aft... | {
"decision": "Reject"
} |
57EjN072hl | 2406.12140v1 | COT Flow: Learning Optimal-Transport Image Sampling and Editing by Contrastive Pairs | {
"content": "## Abstract\n\nAbstract Diffusion models have demonstrated strong performance in sampling and editing multi-modal data with high generation quality, yet they suffer from the iterative generation process which is computationally expensive and slow. In addition, most methods are constrained to generate da... | [
{
"id": "vH15vTT6mK",
"initial_rating": 5,
"confidence": 3,
"soundness": 2,
"contribution": 2,
"presentation": 2,
"summary": "This paper present Contrastive Optimal Transport Flow (COT Flow), a method that achieves fast and high-quality generation with improved zero-shot editing flexibil... | {
"rating": "3;5;5",
"rating_avg": 4.333333333333333,
"confidence": "4;4;3",
"confidence_avg": 3.6666666666666665,
"soundness": "2;3;2",
"soundness_avg": 2.3333333333333335,
"contribution": "2;3;2",
"contribution_avg": 2.3333333333333335,
"presentation": "2;2;2",
"presentation_avg": 2
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:00.844661"
} | {
"id": "Kzz883aG6L",
"metareview": "The paper introduces the Contrastive Optimal Transport Flow (COT Flow), aimed at improving the speed and quality of image generation, with a focus on zero-shot editing flexibility. The model integrates neural optimal transport with consistency models for defining positive pairs ... | {
"decision": "Reject"
} |
57iQSl2G2Q | 2408.16307v1 | Safe Bayesian Optimization for Complex Control Systems via Additive Gaussian Processes | {
"content": "## Abstract\n\nAbstract Controller tuning and optimization have been among the most fundamental problems in robotics and mechatronic systems. The traditional methodology is usually model-based, but its performance heavily relies on an accurate mathematical model of the system. In control applications wi... | [
{
"id": "BR6KedqegB",
"initial_rating": 5,
"confidence": 2,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "In this paper, the authors propose a safe Bayesian optimization framework utilizing Gaussian processes with additive squared exponential functions. The motivation... | {
"rating": "1;3;5;10",
"rating_avg": 4.75,
"confidence": "4;5;2;4",
"confidence_avg": 3.75,
"soundness": "2;2;3;4",
"soundness_avg": 2.75,
"contribution": "2;2;2;4",
"contribution_avg": 2.5,
"presentation": "2;3;3;4",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:00.845302"
} | {
"id": "dHe5kHtMnf",
"metareview": "This paper investigate a SafeControlBO method to optimize multiple controllers at the same time. It shows improvements addressing the challenges associated with tuning complex control systems that involve multiple parameters while keeping online safety.\nStrengths of the paper: ... | {
"decision": "Reject"
} |
58AhfT4Zz1 | 2405.16489v1 | Causal-aware Graph Neural Architecture Search under Distribution Shifts | {
"content": "## Abstract\n\nAbstract Graph neural architecture search (Graph NAS) has emerged as a promising approach for autonomously designing graph neural network architectures by leveraging the correlations between graphs and architectures. However, the existing methods fail to generalize under distribution shif... | [
{
"id": "bibIZ4uDxI",
"initial_rating": 6,
"confidence": 3,
"soundness": 2,
"contribution": 2,
"presentation": 3,
"summary": "The paper proposes a novel method, Causal-aware Graph Neural Architecture Search (CARNAS), to enhance the generalizability of Graph Neural Network (GNN) architect... | {
"rating": "3;5;6;6",
"rating_avg": 5,
"confidence": "5;4;3;3",
"confidence_avg": 3.75,
"soundness": "1;3;3;2",
"soundness_avg": 2.25,
"contribution": "2;2;3;2",
"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": "2026-01-14T22:16:00.846087"
} | {
"id": "MDC9BtHy4H",
"metareview": "This paper proposes a causal-aware graph NAS framework to address the challenge of distribution shifts. By leveraging causal relationships between graph structures and architectures, the method aims to mitigate reliance on spurious correlations and enhance out-of-distribution ge... | {
"decision": "Reject"
} |
5AtlfHYCPa | 2409.18885v1 | HR-Extreme: A High-Resolution Dataset for Extreme Weather Forecasting | {
"content": "## Abstract\n\nAbstract The application of large deep learning models in weather forecasting has led to significant advancements in the field, including higher-resolution forecasting and extended prediction periods exemplified by models such as Pangu and Fuxi. Despite these successes, previous research ... | [
{
"id": "r389JZ1OuR",
"initial_rating": 5,
"confidence": 5,
"soundness": 2,
"contribution": 2,
"presentation": 3,
"summary": "Extreme weather forecasting is a crucial problem for the whole world. With the rise of deep learning-based weather forecasting models, the effectiveness of them o... | {
"rating": "3;5;6;8",
"rating_avg": 5.5,
"confidence": "4;5;4;4",
"confidence_avg": 4.25,
"soundness": "2;2;2;3",
"soundness_avg": 2.25,
"contribution": "2;2;3;3",
"contribution_avg": 2.5,
"presentation": "3;3;2;2",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:00.847081"
} | {
"id": "ZVexEHmt3W",
"metareview": "The authors introduce a high-resolution weather dataset for extreme weather based on the HRRR public data from NOAA (numerical weather prediction). They do this by processing HRRR through unsupervised clustering and filtering and creating a comprehensive benchmark of many extrem... | {
"decision": "Accept (Poster)"
} |
5BjQOUXq7i | 2407.01492v1 | RegMix: Data Mixture as Regression for Language Model Pre-training | {
"content": "## Abstract\n\nAbstract The data mixture for large language model pre-training significantly impacts performance, yet how to determine an effective mixture remains unclear. We propose RegMix to automatically identify a high-performing data mixture by formulating it as a regression task. RegMix involves ... | [
{
"id": "zpE2b0tFPF",
"initial_rating": 8,
"confidence": 4,
"soundness": 4,
"contribution": 4,
"presentation": 4,
"summary": "The work introduces REGMIX, a method for optimizing data mixtures to enhance language model training efficiency. REGMIX treats data mixture selection as a regress... | {
"rating": "5;6;8;8;8",
"rating_avg": 7,
"confidence": "3;4;3;3;4",
"confidence_avg": 3.4,
"soundness": "2;3;3;3;4",
"soundness_avg": 3,
"contribution": "2;3;3;4;4",
"contribution_avg": 3.2,
"presentation": "3;3;3;4;4",
"presentation_avg": 3.4
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Spotlight",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:00.847756"
} | {
"id": "Gvm4ApKQsK",
"metareview": "This paper introduces REGMIX, a novel method for optimizing data mixtures to improve the efficiency of pre-training LLMs. The method frames data mixture selection as a regression task, using small proxy models trained on diverse data mixtures to predict the best-performing mixtu... | {
"decision": "Accept (Spotlight)"
} |
5DT0t5NylU | 2410.00255v1 | Robin3D: Improving 3D Large Language Model via Robust Instruction Tuning | {
"content": "## Abstract\n\nAbstract Recent advancements in 3D Large Language Models (3DLLMs) have highlighted their potential in building general-purpose agents in the 3D real world,\nyet challenges remain due to the lack of high-quality robust instruction-following data, leading to limited discriminative power and... | [
{
"id": "lyGpw8ZaUk",
"initial_rating": 5,
"confidence": 5,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "This paper introduces Robin3D, a 3D large language model trained to follow instructions in 3D environments using the Robust Instruction Generation (RIG) engine, w... | {
"rating": "5;5;5;6;6",
"rating_avg": 5.4,
"confidence": "3;4;5;3;3",
"confidence_avg": 3.6,
"soundness": "3;2;3;3;3",
"soundness_avg": 2.8,
"contribution": "2;2;2;4;3",
"contribution_avg": 2.6,
"presentation": "3;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": "2026-01-14T22:16:00.848403"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
5DUekOKWcS | 2404.08003v3 | Asynchronous Federated Reinforcement Learning with Policy Gradient Updates: Algorithm Design and Convergence Analysis | {
"content": "## Abstract\n\nAbstract To improve the efficiency of reinforcement learning (RL), we propose a novel asynchronous federated reinforcement learning (FedRL) framework termed AFedPG, which constructs a global model through collaboration among N 𝑁 N italic_N agents using policy gradient (PG) updates. To ad... | [
{
"id": "yf4s2H1Kzz",
"initial_rating": 5,
"confidence": 3,
"soundness": 3,
"contribution": 2,
"presentation": 2,
"summary": "The paper aims to enhance the efficiency of federated reinforcement learning (FedRL) by introducing an asynchronous framework, AFedPG, which leverages policy grad... | {
"rating": "3;5;5;5;6",
"rating_avg": 4.8,
"confidence": "4;3;3;3;4",
"confidence_avg": 3.4,
"soundness": "2;3;3;3;3",
"soundness_avg": 2.8,
"contribution": "2;3;2;2;3",
"contribution_avg": 2.4,
"presentation": "2;3;3;2;2",
"presentation_avg": 2.4
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:00.849154"
} | {
"id": "B2AHFuyGBa",
"metareview": "This paper introduces AFedPG, an asynchronous framework for federated reinforcement learning (FedRL), which utilizes policy gradient (PG) updates from multiple agents without requiring synchronized updates. The approach is specifically designed to address challenges in federated... | {
"decision": "Accept (Poster)"
} |
5EuAMDMPRK | 2410.12999v1 | POROver: Improving Safety and Reducing Overrefusal in Large Language Models with Overgeneration and Preference Optimization | {
"content": "## Abstract\n\nAbstract Balancing safety and usefulness in large language models has become a critical challenge in recent years.\nModels often exhibit unsafe behavior or adopt an overly cautious approach, leading to frequent overrefusal of benign prompts, which reduces their usefulness.\nAddressing the... | [
{
"id": "k8ldmVZYPP",
"initial_rating": 6,
"confidence": 3,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "The authors present a framework that aims to reduce overrefusal in Large Language Models (LLMs), while improving their safety. It involves finetuning on overgener... | {
"rating": "3;6;6",
"rating_avg": 5,
"confidence": "4;4;3",
"confidence_avg": 3.6666666666666665,
"soundness": "2;2;3",
"soundness_avg": 2.3333333333333335,
"contribution": "2;2;3",
"contribution_avg": 2.3333333333333335,
"presentation": "2;2;4",
"presentation_avg": 2.6666666666666665
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:00.849960"
} | {
"id": "2VdJIaJBOc",
"metareview": "This paper aims to understand the beenefits of over generation, and forms of finetuning/rejection sampling to improve usefulness/safety tradeoffs in LLMs. While authors agreed that this is an important direction of study, many felt that it was somewhat incremental compared to pr... | {
"decision": "Reject"
} |
5GZuEZDmUE | 2405.17823v3 | Spectral Truncation Kernels: Noncommutativity in $C^*$-algebraic Kernel Machines | {
"content": "## Abstract\n\nAbstract C ∗ superscript 𝐶 C^{*} italic_C start_POSTSUPERSCRIPT ∗ end_POSTSUPERSCRIPT -algebra-valued kernels could pave the way for the next generation of kernel machines. To further our fundamental understanding of learning with C ∗ superscript 𝐶 C^{*} italic_C start_POSTSUPERSCRIPT ∗... | [
{
"id": "ZkczZnLKtN",
"initial_rating": 3,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "This paper explores the recent subfield of positive definite kernels with values in a C*-algebra and RKHM (the correponding \"RKHS\" theory) . The whole work is m... | {
"rating": "3;3;5;5",
"rating_avg": 4,
"confidence": "3;4;2;3",
"confidence_avg": 3,
"soundness": "3;3;3;2",
"soundness_avg": 2.75,
"contribution": "2;2;3;2",
"contribution_avg": 2.25,
"presentation": "2;2;3;3",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:00.850750"
} | {
"id": "wbXxGF6Z55",
"metareview": "This paper utilizes the tool of C-algebra to develop new methods of kernel machines, receiving certain recognition from the reviewers. However, the current paper suffers from obvious weaknesses that prevent me from recommending acceptance. To me specific, the experimental setup,... | {
"decision": "Reject"
} |
5GgjiRzYp3 | 2405.18295v2 | Intent3D: 3D Object Detection in RGB-D Scans Based on Human Intention | {
"content": "## Abstract\n\nAbstract In real-life scenarios, humans seek out objects in the 3D world to fulfill their daily needs or intentions. This inspires us to introduce 3D intention grounding, a new task in 3D object detection employing RGB-D, based on human intention, such as \" I want something to support my... | [
{
"id": "avUA49V4t6",
"initial_rating": 5,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "This paper introduces a new task named 3D-intention grounding, which is 3D object-detection from direct human-intentions. The paper collects Intent3D dataset whic... | {
"rating": "5;5;5;6",
"rating_avg": 5.25,
"confidence": "2;2;3;4",
"confidence_avg": 2.75,
"soundness": "3;2;3;3",
"soundness_avg": 2.75,
"contribution": "3;2;3;3",
"contribution_avg": 2.75,
"presentation": "3;2;3;2",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:00.851609"
} | {
"id": "5bHfaW26Pn",
"metareview": "This paper introduces the 3D Intention Grounding (3D-IG) task, which aims to detect objects in 3D scenes based on human intention instructions. To support this task, the authors develop the Intent3D dataset and propose a method called IntentNet, which consists of three modules: ... | {
"decision": "Accept (Poster)"
} |
5GuhYMgaap | 2408.00114v2 | Inductive or Deductive? Rethinking the Fundamental Reasoning Abilities of LLMs | {
"content": "## Abstract\n\nAbstract Reasoning encompasses two typical types: deductive reasoning and inductive reasoning. Despite extensive research into the reasoning capabilities of Large Language Models (LLMs), most studies have failed to rigorously differentiate between inductive and deductive reasoning, leadin... | [
{
"id": "ApOqTTs00x",
"initial_rating": 5,
"confidence": 4,
"soundness": 3,
"contribution": 2,
"presentation": 2,
"summary": "This paper studies LLM capabilities in inductive and deductive reasoning, and compares the performance gap between the two poles of reasoning. They consider this ... | {
"rating": "3;5;5;5",
"rating_avg": 4.5,
"confidence": "4;3;3;4",
"confidence_avg": 3.5,
"soundness": "2;2;3;3",
"soundness_avg": 2.5,
"contribution": "1;2;3;2",
"contribution_avg": 2,
"presentation": "3;2;4;2",
"presentation_avg": 2.75
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:00.852239"
} | {
"id": "qIZZLbKJsU",
"metareview": "All reviewers agreed that this paper potentially provides a novel way to understand the interaction with LLMs. The set of benchmarks is varied, and the paper is rather well-written. \n\nHowever, the novelty should be made clearer, and the work should provide a better placement w... | {
"decision": "Reject"
} |
5IZfo98rqr | 2410.14670v1 | Decomposing The Dark Matter of Sparse Autoencoders | {
"content": "## Abstract\n\nAbstract Sparse autoencoders (SAEs) are a promising technique for decomposing language model activations into interpretable linear features. However, current SAEs fall short of completely explaining model performance, resulting in “dark matter”: unexplained variance in activations. This w... | [
{
"id": "38zcCt6YBA",
"initial_rating": 3,
"confidence": 4,
"soundness": 2,
"contribution": 3,
"presentation": 4,
"summary": "This paper is an analysis into the error of sparse autoencoders applied to LLM interpredability. They address the shortcoming of the ability to reconstruct the hi... | {
"rating": "3;3;3;5",
"rating_avg": 3.5,
"confidence": "2;2;4;2",
"confidence_avg": 2.5,
"soundness": "2;2;2;2",
"soundness_avg": 2,
"contribution": "3;2;3;3",
"contribution_avg": 2.75,
"presentation": "2;1;4;1",
"presentation_avg": 2
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:00.853143"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
5IkDAfabuo | 2410.18082v1 | Prioritized Generative Replay | {
"content": "## Abstract\n\nAbstract Sample-efficient online reinforcement learning often uses replay buffers to store experience for reuse when updating the value function.\nHowever, uniform replay is inefficient, since certain classes of transitions can be more relevant to learning. While prioritization of more us... | [
{
"id": "h0XUtHBAKU",
"initial_rating": 8,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "The work proposes a form of sample based experience replays that leverages a generative model to provide and augment samples drawn from the replay buffer. To avoi... | {
"rating": "5;5;6;8",
"rating_avg": 6,
"confidence": "3;4;4;3",
"confidence_avg": 3.5,
"soundness": "3;3;4;4",
"soundness_avg": 3.5,
"contribution": "3;2;3;4",
"contribution_avg": 3,
"presentation": "3;3;3;3",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Oral",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:00.853933"
} | {
"id": "4PKgx2c6cC",
"metareview": "This paper uses generative models for samples for experience replay in RL. The paper uses conditional diffusion models, and explores different relevance functions, showing empirical improvements across experiments in online RL. \n\nReviewers agree that this paper is well-written... | {
"decision": "Accept (Oral)"
} |
5KqveQdXiZ | 2410.22796v1 | Solving Differential Equations with Constrained Learning | {
"content": "## Abstract\n\nAbstract (Partial) differential equations (PDEs) are fundamental tools for describing natural phenomena, making their solution crucial in science and engineering. While traditional methods, such as the finite element method, provide reliable solutions, their accuracy is often tied to the ... | [
{
"id": "RIpTc7j2cK",
"initial_rating": 3,
"confidence": 4,
"soundness": 2,
"contribution": 2,
"presentation": 4,
"summary": "This paper aims to adapt ideas from adversarial robustness and use them to find neural network solutions to partial differential equations with improved guarantee... | {
"rating": "3;3;5;8",
"rating_avg": 4.75,
"confidence": "4;4;3;3",
"confidence_avg": 3.5,
"soundness": "1;2;3;3",
"soundness_avg": 2.25,
"contribution": "1;2;2;3",
"contribution_avg": 2,
"presentation": "1;4;3;3",
"presentation_avg": 2.75
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:00.854718"
} | {
"id": "Z4h5UPs1pV",
"metareview": "This work shows that obtaining weak solutions to PDEs can be formulated as a constrained learning problem with worst case losses. Furthermore, it develops an efficient algorithm for solving this problem and incorporating further structural information that maybe be known. The me... | {
"decision": "Accept (Poster)"
} |
5LXcoDtNyq | 2410.11996v1 | Holistic Reasoning with Long-Context LMs: A Benchmark for Database Operations on Massive Textual Data | {
"content": "## Abstract\n\nAbstract The rapid increase in textual information means we need more efficient methods to sift through, organize, and understand it all. While retrieval-augmented generation (RAG) models excel in accessing information from large document collections, they struggle with complex tasks that... | [
{
"id": "IUk141vszU",
"initial_rating": 6,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 4,
"summary": "This paper presents HoloBench, a benchmark designed to evaluate the holistic reasoning capabilities of Long-Context Language Models (LCLMs). While Retrieval-Augme... | {
"rating": "5;5;6;8",
"rating_avg": 6,
"confidence": "4;5;3;4",
"confidence_avg": 4,
"soundness": "2;2;3;4",
"soundness_avg": 2.75,
"contribution": "2;2;3;4",
"contribution_avg": 2.75,
"presentation": "3;3;4;4",
"presentation_avg": 3.5
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:00.855592"
} | {
"id": "QUdqpxhhgd",
"metareview": "In this paper, the authors propose a benchmark for evaluating the capabilities of long-context language models. Reviewers consider this a valuable contribution to the community, given the release of the associated code and data, which facilitate comparisons between long-context ... | {
"decision": "Accept (Poster)"
} |
5MBUmj5mTI | 2410.14878v1 | On the Influence of Shape, Texture and Color for Learning Semantic Segmentation | {
"content": "## Abstract\n\nAbstract In recent years, a body of works has emerged, studying shape and texture biases of off-the-shelf pre-trained deep neural networks (DNN) for image classification. These works study how much a trained DNN relies on image cues, predominantly shape and texture. In this work, we switc... | [
{
"id": "NeFZFZcvxf",
"initial_rating": 5,
"confidence": 4,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "The paper presents an analysis of the influence of shape, texture, and color on semantic segmentation performance, proposing a methodology that leverages augmente... | {
"rating": "5;5;6;6",
"rating_avg": 5.5,
"confidence": "4;4;3;3",
"confidence_avg": 3.5,
"soundness": "2;3;3;3",
"soundness_avg": 2.75,
"contribution": "2;2;3;2",
"contribution_avg": 2.25,
"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": "2026-01-14T22:16:00.856240"
} | {
"id": "MVP0HcFopD",
"metareview": "This paper studies the importance of different image cues like color, shape, texture, etc., for the learning of deep semantic segmentation models. The manuscript was reviewed by four experts in the field. The recommendations are (2 x \"5: marginally below the acceptance threshol... | {
"decision": "Reject"
} |
5Qxx5KpFms | 2409.05780v1 | Breaking Neural Network Scaling Laws with Modularity | {
"content": "## Abstract\n\nAbstract Modular neural networks outperform nonmodular neural networks on tasks ranging from visual question answering to robotics. These performance improvements are thought to be due to modular networks’ superior ability to model the compositional and combinatorial structure of real-wor... | [
{
"id": "HqILmnbtZd",
"initial_rating": 3,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 2,
"summary": "This paper (1) constructs a simplified theoretical model of generalization (focused on the case of linear regression from what I believe to be a set of not-strict... | {
"rating": "3;3;8;8",
"rating_avg": 5.5,
"confidence": "4;3;4;2",
"confidence_avg": 3.25,
"soundness": "2;3;4;4",
"soundness_avg": 3.25,
"contribution": "2;2;3;3",
"contribution_avg": 2.5,
"presentation": "2;2;4;4",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:00.857017"
} | {
"id": "8mNsRZeAJr",
"metareview": "This paper derives a theoretical quantitative analysis of the generalization abilities of modular architectures. The analysis shows that under certain assumptions (linearity, specific modular structure of architecture and data) models can avoid exponential sample complexity asso... | {
"decision": "Accept (Poster)"
} |
5UQ0YmC2js | 2410.21471v2 | AdvI2I: Adversarial Image Attack on Image-to-Image Diffusion models | {
"content": "## Abstract\n\nAbstract Recent advances in diffusion models have significantly enhanced the quality of image synthesis, yet they have also introduced serious safety concerns, particularly the generation of Not Safe for Work (NSFW) content. Previous research has demonstrated that adversarial prompts can ... | [
{
"id": "8ro1VVUrtG",
"initial_rating": 5,
"confidence": 4,
"soundness": 2,
"contribution": 2,
"presentation": 3,
"summary": "This paper introduces AdvI2I, a framework that performs adversarial image attacks on image-to-image (I2I) diffusion models to generate NSFW content without modify... | {
"rating": "3;5;5;6",
"rating_avg": 4.75,
"confidence": "3;5;4;4",
"confidence_avg": 4,
"soundness": "2;2;2;3",
"soundness_avg": 2.25,
"contribution": "2;2;2;3",
"contribution_avg": 2.25,
"presentation": "2;2;3;3",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:00.857935"
} | {
"id": "p090JjwoH7",
"metareview": "This paper studies the vulnerabilities of image-to-image (I2I) diffusion models under adversarial image attacks. A new method AdvI2I is proposed to generate adversarial perturbations on input images to induce NSFW content. The proposed method is effective to mislead Stable Diffu... | {
"decision": "Reject"
} |
5VK1UulEbE | 2410.01860v4 | FredNormer: Frequency Domain Normalization for Non-stationary Time Series Forecasting | {
"content": "## Abstract\n\nAbstract Recent normalization-based methods have shown great success in tackling the distribution shift issue, facilitating non-stationary time series forecasting.\nSince these methods operate in the time domain, they may fail to fully capture the dynamic patterns that are more apparent i... | [
{
"id": "c3u62sfWpU",
"initial_rating": 3,
"confidence": 5,
"soundness": 2,
"contribution": 2,
"presentation": 2,
"summary": "This paper reveals that traditional time-domain normalization methods uniformly scale non-zero frequencies, which limits their ability to effectively handle distr... | {
"rating": "3;3;5;5;5",
"rating_avg": 4.2,
"confidence": "3;5;5;4;4",
"confidence_avg": 4.2,
"soundness": "2;2;2;3;3",
"soundness_avg": 2.4,
"contribution": "2;2;3;2;2",
"contribution_avg": 2.2,
"presentation": "2;2;3;3;3",
"presentation_avg": 2.6
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:00.858534"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
5WtovCb1ZE | 2405.15722v2 | Models That Prove Their Own Correctness | {
"content": "## Abstract\n\nAbstract How can we trust the correctness of a learned model on a particular input of interest? Model accuracy is typically measured on average over a distribution of inputs, giving no guarantee for any fixed input.\nThis paper proposes a theoretically-founded solution to this problem: to... | [
{
"id": "kaiAuvvrjC",
"initial_rating": 8,
"confidence": 4,
"soundness": 3,
"contribution": 4,
"presentation": 4,
"summary": "In this paper, the authors propose a new type of self-proving models that not just predict an output for a given input but also a proof for the correctness of the... | {
"rating": "1;3;6;8",
"rating_avg": 4.5,
"confidence": "1;3;3;4",
"confidence_avg": 2.75,
"soundness": "3;2;3;3",
"soundness_avg": 2.75,
"contribution": "3;2;3;4",
"contribution_avg": 3,
"presentation": "3;2;3;4",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:00.859333"
} | {
"id": "KKMUM1KQoy",
"metareview": "The paper introduces a framework when ML models can prove the correctness of their output. The notion of proof here is \"interactive proofs\" as studied in computational complexity, with a polynomially-bounded verifier and a potentially unbounded prover that can interact. When t... | {
"decision": "Reject"
} |
5ZkuWAbxzT | 2411.04625v1 | Sharp Analysis for KL-Regularized Contextual Bandits and RLHF | {
"content": "## Abstract\n\nAbstract Reverse-Kullback-Leibler (KL) regularization has emerged to be a predominant technique used to enhance policy optimization in reinforcement learning (RL) and reinforcement learning from human feedback (RLHF), which forces the learned policy to stay close to a reference policy. Wh... | [
{
"id": "BD15zhtwBv",
"initial_rating": 6,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "In this paper, the authors provide a new analysis of contextual bandits under KL regularization that achieves an improved sample complexity guarantee. Then, they ... | {
"rating": "3;3;5;6",
"rating_avg": 4.25,
"confidence": "4;2;3;3",
"confidence_avg": 3,
"soundness": "3;2;3;3",
"soundness_avg": 2.75,
"contribution": "2;2;2;3",
"contribution_avg": 2.25,
"presentation": "2;2;3;3",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:00.861052"
} | {
"id": "NKFwdU7S3l",
"metareview": "This work obtains a tight bound on the sample complexity of RLHF with KL regularization where they show that the sample complexity scales with 1/epsilon instead of 1/epsilon^2. Unfortunately a couple of weaknesses were pointed out by the reviewers including the strength of the c... | {
"decision": "Reject"
} |
5ck9PIrTpH | 2410.13502v2 | MathGAP: Out-of-Distribution Evaluation on Problems with Arbitrarily Complex Proofs | {
"content": "## Abstract\n\nAbstract Large language models (LLMs) can solve arithmetic word problems with high accuracy, but little is known about how well they generalize to problems that are more complex than the ones on which they have been trained. Empirical investigations of such questions are impeded by two ma... | [
{
"id": "G2jRt0YPwB",
"initial_rating": 8,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "This paper introduces MathGAP, a framework designed to evaluate large language models (LLMs) on mathematical word problems requiring proofs of arbitrary complexit... | {
"rating": "5;5;6;8",
"rating_avg": 6,
"confidence": "3;5;3;3",
"confidence_avg": 3.5,
"soundness": "3;3;3;3",
"soundness_avg": 3,
"contribution": "2;2;3;3",
"contribution_avg": 2.5,
"presentation": "4;4;3;3",
"presentation_avg": 3.5
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:00.862067"
} | {
"id": "ZbKcSi9g6c",
"metareview": "(a) Scientific Claims and Findings:\nThe paper introduces MathGAP, a framework for evaluating LLMs on arithmetic word problems with arbitrarily complex proofs. Key findings include:\n- Most tested LLMs show significant performance degradation as proof complexity increases, parti... | {
"decision": "Accept (Poster)"
} |
5dDYhvt6dY | 2410.04731v1 | Efficient transformer with reinforced position embedding for language models | {
"content": "## Abstract\n\nAbstract In this paper, we propose an efficient transformer architecture that uses reinforced positional embedding to obtain superior performance with half the number of encoder decoder layers. We demonstrate that concatenating positional encoding with trainable token embeddings, normaliz... | [
{
"id": "RTfdRmZLEp",
"initial_rating": 3,
"confidence": 3,
"soundness": 2,
"contribution": 2,
"presentation": 3,
"summary": "This paper studies how to make transformer architecture more efficient by reinforcing the positional embedding. To achieve this goal, the paper makes three propos... | {
"rating": "3;3;3;5",
"rating_avg": 3.5,
"confidence": "5;4;3;4",
"confidence_avg": 4,
"soundness": "2;2;2;2",
"soundness_avg": 2,
"contribution": "2;2;2;3",
"contribution_avg": 2.25,
"presentation": "2;2;3;3",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:00.862669"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
5dttvRONu0 | 2410.04661v1 | Federated Learning Nodes Can Reconstruct Peers' Image Data | {
"content": "## Abstract\n\nAbstract Federated learning (FL) is a privacy-preserving machine learning framework that enables multiple nodes to train models on their local data and periodically average weight updates to benefit from other nodes’ training. Each node’s goal is to collaborate with other nodes to improve... | [
{
"id": "bB46WSocl0",
"initial_rating": 3,
"confidence": 4,
"soundness": 2,
"contribution": 2,
"presentation": 3,
"summary": "This paper proposes an attack approach within the federated learning (FL) framework to reconstruct image data from participating peers in a centralized system. Th... | {
"rating": "3;3;5",
"rating_avg": 3.6666666666666665,
"confidence": "5;4;5",
"confidence_avg": 4.666666666666667,
"soundness": "2;2;3",
"soundness_avg": 2.3333333333333335,
"contribution": "1;2;2",
"contribution_avg": 1.6666666666666667,
"presentation": "2;3;3",
"presentation_avg": 2.66666666666666... | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:00.863171"
} | {
"id": "GOZffo1u5y",
"metareview": "The paper proposes and demonstrates a gradient inversion attack in federated learning by a client node against other client nodes.\nDemonstrating an attack in a novel setting is potentially very interesting.\nThe reviewers criticise the paper for unrealistic assumptions and limi... | {
"decision": "Reject"
} |
5f3brwjeTl | 2404.12916v2 | Physical Backdoor Attack can Jeopardize Driving with Vision-Large-Language Models | {
"content": "## Abstract\n\nAbstract Vision-Large-Language-models (VLMs) have great application prospects in autonomous driving.\nDespite the ability of VLMs to comprehend and make decisions in complex scenarios, their integration into safety-critical autonomous driving systems poses serious security risks.\nIn this... | [
{
"id": "caBBvblBV6",
"initial_rating": 3,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "This paper introduces BadVLMDriver, the first physical backdoor attack targeting vision large language models (VLMs) in autonomous driving. Using everyday objects... | {
"rating": "5;5;6;6",
"rating_avg": 5.5,
"confidence": "4;3;3;3",
"confidence_avg": 3.25,
"soundness": "2;3;3;3",
"soundness_avg": 2.75,
"contribution": "2;3;2;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": "2026-01-14T22:16:00.863776"
} | {
"id": "TnEiqMHU63",
"metareview": "This paper proposes BadVLMDriver, a backdoor attack method against VLMs for autonomous driving. To enhance practicality, the authors use common physical objects (a red balloon), to initiate unsafe actions like sudden acceleration, highlighting a real-world threat to autonomous v... | {
"decision": "Reject"
} |
5lIXRf8Lnw | 2410.13928v1 | Automatically Interpreting Millions of Features in Large Language Models | {
"content": "## Abstract\n\nAbstract While the activations of neurons in deep neural networks usually do not have a simple human-understandable interpretation, sparse autoencoders (SAEs) can be used to transform these activations into a higher-dimensional latent space which may be more easily interpretable. However,... | [
{
"id": "GS5FxphDzr",
"initial_rating": 3,
"confidence": 4,
"soundness": 2,
"contribution": 3,
"presentation": 3,
"summary": "This paper introduced five automated scoring methods to score the explanations of SAE latents, and discussed the shortcomings of existing scoring techniques. The ... | {
"rating": "3;3;5;8",
"rating_avg": 4.75,
"confidence": "3;4;4;4",
"confidence_avg": 3.75,
"soundness": "2;2;2;4",
"soundness_avg": 2.5,
"contribution": "2;3;3;4",
"contribution_avg": 3,
"presentation": "1;3;2;3",
"presentation_avg": 2.25
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:00.864479"
} | {
"id": "N5hCx39mMT",
"metareview": "## Summary of Scientific Claims and Findings \nThe paper introduces a novel automated framework using large language models (LLMs) to generate and evaluate natural language explanations for Sparse Autoencoder (SAE) features. It proposes five new scoring methods—detection, fuzzi... | {
"decision": "Reject"
} |
5m43PEd3sz | 2410.05225v1 | ETGL-DDPG: A Deep Deterministic Policy Gradient Algorithm for Sparse Reward Continuous Control | {
"content": "## Abstract\n\nAbstract We consider deep deterministic policy gradient (DDPG) in the context of reinforcement learning with sparse rewards. To enhance exploration, we introduce a search procedure, ϵ t italic-ϵ 𝑡 {\\epsilon}{t} italic_ϵ italic_t -greedy , which generates exploratory options for explor... | [
{
"id": "D90bOCVOYP",
"initial_rating": 5,
"confidence": 2,
"soundness": 3,
"contribution": 3,
"presentation": 2,
"summary": "This paper addresses the limitations of the DDPG algorithm in sparse-reward environments. It identifies three main deficiencies: lack of directional exploration, ... | {
"rating": "3;3;5;5",
"rating_avg": 4,
"confidence": "3;4;4;2",
"confidence_avg": 3.25,
"soundness": "2;2;3;3",
"soundness_avg": 2.5,
"contribution": "2;2;2;3",
"contribution_avg": 2.25,
"presentation": "3;3;3;2",
"presentation_avg": 2.75
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:00.865242"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
5ncdKonxd4 | 2410.17247v1 | PyramidDrop: Accelerating Your Large Vision-Language Models via Pyramid Visual Redundancy Reduction | {
"content": "## Abstract\n\nAbstract In large vision-language models (LVLMs), images serve as inputs that carry a wealth of information. As the idiom “A picture is worth a thousand words” implies, representing a single image in current LVLMs can require hundreds or even thousands of tokens. This results in significa... | [
{
"id": "s1kPo9Pfsf",
"initial_rating": 3,
"confidence": 5,
"soundness": 2,
"contribution": 2,
"presentation": 3,
"summary": "This paper introduces PyramidDrop, a method to improve the efficiency of Large Vision-Language Models by progressively reducing visual tokens across model layers.... | {
"rating": "3;3;3;5",
"rating_avg": 3.5,
"confidence": "4;5;5;4",
"confidence_avg": 4.5,
"soundness": "3;3;2;3",
"soundness_avg": 2.75,
"contribution": "2;2;2;3",
"contribution_avg": 2.25,
"presentation": "2;2;3;3",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:00.865969"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
5o0phqAhsP | 2402.04398v1 | Learning under Temporal Label Noise | {
"content": "## Abstract\n\nAbstract Many sequential classification tasks are affected by label noise that varies over time.\nSuch noise can cause label quality to improve, worsen, or periodically change over time. We first propose and formalize temporal label noise , an unstudied problem for sequential classificati... | [
{
"id": "b81m3MeHNH",
"initial_rating": 6,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 4,
"summary": "The manuscript introduced temporal label noise in time series classification tasks and proposed a novel framework that are robust to it. Experiments were conducte... | {
"rating": "3;5;6;8",
"rating_avg": 5.5,
"confidence": "3;3;4;3",
"confidence_avg": 3.25,
"soundness": "2;2;3;4",
"soundness_avg": 2.75,
"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 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:00.866565"
} | {
"id": "nxa2rjMCwf",
"metareview": "A fairly good paper that should be accepted for publication at ICLR. I hope this paper can advance the research area label-noise learning.\n\nMy only comment is about related work. The standard non-temporal label noise can be regarded as a type of distribution shift, where the t... | {
"decision": "Accept (Poster)"
} |
5o9JJJPPm6 | 2410.01954v1 | ComaDICE: Offline Cooperative Multi-Agent Reinforcement Learning with Stationary Distribution Shift Regularization | {
"content": "## Abstract\n\nAbstract Offline reinforcement learning (RL) has garnered significant attention for its ability to learn effective policies from pre-collected datasets without the need for further environmental interactions. While promising results have been demonstrated in single-agent settings, offline... | [
{
"id": "oGBt47piUj",
"initial_rating": 6,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 2,
"summary": "This paper proposes an algorithm introducing stationary distribution correction to address the distributional shift problem in offline cooperative multi-agent rei... | {
"rating": "3;3;6;6",
"rating_avg": 4.5,
"confidence": "4;5;4;3",
"confidence_avg": 4,
"soundness": "3;2;3;3",
"soundness_avg": 2.75,
"contribution": "1;1;3;3",
"contribution_avg": 2,
"presentation": "3;1;2;2",
"presentation_avg": 2
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:00.867281"
} | {
"id": "lNg426iiVv",
"metareview": "The paper studied offline multi-agent RL and proposed an approach based on the DICE framework. The proposed approach uses a stationary distribution shift regularization to combat the distribution shift issue in offline RL. The paper demonstrates that their approach works well em... | {
"decision": "Accept (Poster)"
} |
5oRB2Wgwtb | 2410.03230v1 | Online Bandit Nonlinear Control with Dynamic Batch Length and Adaptive Learning Rate | {
"content": "## Abstract\n\nAbstract This paper is concerned with the online bandit nonlinear control, which aims to learn the best stabilizing controller from a pool of stabilizing and destabilizing controllers of unknown types for a given nonlinear dynamical system. We develop an algorithm, named D ynamic B atch l... | [
{
"id": "4493rZkYBi",
"initial_rating": 5,
"confidence": 3,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "The paper addresses the online bandit nonlinear control problem, where the objective is to learn an optimal controller for a nonlinear dynamical system amid unkno... | {
"rating": "5;5;5;6",
"rating_avg": 5.25,
"confidence": "3;3;3;3",
"confidence_avg": 3,
"soundness": "3;3;3;3",
"soundness_avg": 3,
"contribution": "3;2;2;2",
"contribution_avg": 2.25,
"presentation": "3;2;3;2",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:00.868471"
} | {
"id": "PG9gNdLSsG",
"metareview": "Summary:\nThis paper investigates online nonlinear control using bandit feedback. Its primary contribution lies in introducing dynamic batch length and learning rate techniques to relax the commonly used exponentially stabilizing controller assumption. Instead, it adopts a weake... | {
"decision": "Reject"
} |
5pFV1FxG9d | 2410.13331v1 | Improving Discrete Optimisation Via Decoupled Straight-Through Gumbel-Softmax | {
"content": "## Abstract\n\nAbstract Discrete representations play a crucial role in many deep learning architectures, yet their non-differentiable nature poses significant challenges for gradient-based optimization. To address this issue, various gradient estimators have been developed, including the Straight-Throu... | [
{
"id": "1ZuXMRmWsu",
"initial_rating": 3,
"confidence": 3,
"soundness": 1,
"contribution": 2,
"presentation": 2,
"summary": "This paper present a simple method, called decoupled stgs,for dealing with discrete representation. Through the employing the decoupled temperatures for forward a... | {
"rating": "3;3;5;6",
"rating_avg": 4.25,
"confidence": "5;3;3;3",
"confidence_avg": 3.5,
"soundness": "2;1;2;3",
"soundness_avg": 2,
"contribution": "1;2;1;3",
"contribution_avg": 1.75,
"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": "2026-01-14T22:16:00.869517"
} | {
"id": "CoJG7VZ3Ug",
"metareview": "The work empirically improves the ST-GS estimator commonly used in discrete settings by decoupling the temperature parameters used in forward and backward passes. The suitable temperatures are determined based on grid search. The idea is simple and shown to be effective. \n\nThe... | {
"decision": "Reject"
} |
5qg1sAXhoh | 2406.10411v1 | Tree Search for Simultaneous Move Games via Equilibrium Approximation | {
"content": "## Abstract\n\nAbstract Neural network supported tree-search has shown strong results in a variety of perfect information multi-agent tasks. However, the performance of these methods on partial information games has generally been below competing approaches. Here we study the class of simultaneous-move ... | [
{
"id": "vywOG6jhKl",
"initial_rating": 8,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "This paper develops a method that combines deep Monte Carlo Tree Search with online no-regret learning in order to approximate coarse correlated equilibria in bot... | {
"rating": "3;3;5;8",
"rating_avg": 4.75,
"confidence": "5;4;4;3",
"confidence_avg": 4,
"soundness": "2;2;2;3",
"soundness_avg": 2.25,
"contribution": "2;2;2;3",
"contribution_avg": 2.25,
"presentation": "2;2;1;3",
"presentation_avg": 2
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:00.870112"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
5sPgOyyjG5 | 2407.02010v1 | Feynman-Kac Operator Expectation Estimator | {
"content": "## Abstract\n\nAbstract The Feynman-Kac Operator Expectation Estimator (FKEE) is an innovative method for estimating the target Mathematical Expectation 𝔼 X ∼ P [ f ( X ) ] subscript 𝔼 similar-to 𝑋 𝑃 delimited-[] 𝑓 𝑋 \\mathbb{E}_{X\\sim P}[f(X)] blackboard_E start_POSTSUBSCRIPT italic_X ∼ ital... | [
{
"id": "tCbFqq2ZJ6",
"initial_rating": 1,
"confidence": 4,
"soundness": 1,
"contribution": 1,
"presentation": 1,
"summary": "Doing MCMC is hard (time consuming, and somewhat wasteful because of the burn-in period). The authors propose a post-processing method using the samples from some... | {
"rating": "1;1;3;3;5",
"rating_avg": 2.6,
"confidence": "5;4;3;3;4",
"confidence_avg": 3.8,
"soundness": "2;1;3;2;2",
"soundness_avg": 2,
"contribution": "2;1;2;2;2",
"contribution_avg": 1.8,
"presentation": "2;1;2;2;2",
"presentation_avg": 1.8
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:00.870842"
} | {
"id": "DRbnI0hMVq",
"metareview": "**Summary of Discussion:** \nThe reviewers appreciated the novel ideas proposed in the paper, such as using the Feynman-Kac equation (FKE) and Physics-Informed Neural Networks (PINNs) to estimate mathematical expectations. However, the submission has significant issues that hin... | {
"decision": "Reject"
} |
5sQiK2qTGa | 2410.23123v1 | On Memorization of Large Language Models in Logical Reasoning | {
"content": "## Abstract\n\nAbstract Large language models (LLMs) achieve good performance on challenging reasoning benchmarks, yet could also make basic reasoning mistakes. This contrasting behavior is puzzling when it comes to understanding the mechanisms behind LLMs’ reasoning capabilities.\nOne hypothesis is tha... | [
{
"id": "qSTPAs2idC",
"initial_rating": 5,
"confidence": 4,
"soundness": 2,
"contribution": 2,
"presentation": 3,
"summary": "This study examines how LLMs balance memorization and reasoning in solving logical reasoning tasks, using a benchmark based on Knights and Knaves (K&K) puzzles. F... | {
"rating": "3;3;5;5",
"rating_avg": 4,
"confidence": "5;4;3;4",
"confidence_avg": 4,
"soundness": "2;1;3;2",
"soundness_avg": 2,
"contribution": "2;1;2;2",
"contribution_avg": 1.75,
"presentation": "4;2;3;3",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:00.871799"
} | {
"id": "xuwtB6qZE1",
"metareview": "Whether LLMs learn to reason or its perceived reasoning power is rooted in its ability to memorize huge space of potential answers is an important question to understand the mechanism of LLMs. Because of its importance, lots of prior work exists. This paper proposes a \"perturba... | {
"decision": "Reject"
} |
5sdUTpDlbX | 2409.20158v1 | Professor X: Manipulating EEG BCI with Invisible and Robust Backdoor Attack | {
"content": "## Abstract\n\nAbstract While electroencephalogram (EEG) based brain-computer interface (BCI) has been widely used for medical diagnosis, health care, and device control, the safety of EEG BCI has long been neglected.\nIn this paper, we propose Professor X , an invisible and robust “mind-controller” tha... | [
{
"id": "wCssqcfUio",
"initial_rating": 1,
"confidence": 5,
"soundness": 1,
"contribution": 1,
"presentation": 1,
"summary": "From the outset, the abstract of the submission presents a proposition that appears to be unrealistic and somewhat disconnected from contemporary research realiti... | {
"rating": "1;3;5;6;8",
"rating_avg": 4.6,
"confidence": "5;5;3;3;4",
"confidence_avg": 4,
"soundness": "1;2;2;4;4",
"soundness_avg": 2.6,
"contribution": "1;1;2;4;3",
"contribution_avg": 2.2,
"presentation": "1;2;2;4;4",
"presentation_avg": 2.6
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:00.872746"
} | {
"id": "p9cC5WfaUo",
"metareview": "The paper introduces a novel and robust backdoor attack framework for EEG-based Brain-Computer Interfaces. The method employs a label-poisoning strategy with reinforcement learning to inject undetectable triggers into EEG signals. The paper introduces and tackles a very interest... | {
"decision": "Reject"
} |
5tjdRyqnSn | 2405.14650v1 | PhiNets: Brain-inspired Non-contrastive Learning Based on Temporal Prediction Hypothesis | {
"content": "## Abstract\n\nAbstract SimSiam is a prominent self-supervised learning method that achieves impressive results in various vision tasks under static environments.\nHowever, it has two critical issues: high sensitivity to hyperparameters, especially weight decay, and unsatisfactory performance in online ... | [
{
"id": "oiv2efjYKH",
"initial_rating": 6,
"confidence": 3,
"soundness": 2,
"contribution": 2,
"presentation": 3,
"summary": "In this paper, the authors propose PhiNet, a self-supervised learning (SSL) method partly inspired by the temporal prediction hypothesis and a previous SSL method... | {
"rating": "6;6;6;8",
"rating_avg": 6.5,
"confidence": "2;3;3;4",
"confidence_avg": 3,
"soundness": "3;3;2;3",
"soundness_avg": 2.75,
"contribution": "4;3;2;3",
"contribution_avg": 3,
"presentation": "3;1;3;2",
"presentation_avg": 2.25
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:00.873524"
} | {
"id": "K0lLYD43pU",
"metareview": "This paper presents a novel self-supervised learning technique, loosely inspired by work on the hippocampus in neuroscience, called PhiNet. Similar to SimSiam, PhiNet take an input, generates two augmentations. But, unlike PhiNet, it makes a prediction with both augmentations (o... | {
"decision": "Accept (Poster)"
} |
5uUr3WFmyZ | 2406.16649v2 | Almost sure convergence of stochastic Hamiltonian descent methods | {
"content": "## Abstract\n\nAbstract. Gradient normalization and soft clipping are two popular techniques for tackling instability issues and improving convergence of stochastic gradient descent (SGD) with momentum.\nIn this article, we study these types of methods through the lens of dissipative Hamiltonian systems... | [
{
"id": "uvucgat3vm",
"initial_rating": 6,
"confidence": 2,
"soundness": 4,
"contribution": 3,
"presentation": 4,
"summary": "The paper studies a family of stochastic gradient methods given by equation 9 and for this family almost sure convergence to stationary points is proved under the... | {
"rating": "3;6;6",
"rating_avg": 5,
"confidence": "5;4;2",
"confidence_avg": 3.6666666666666665,
"soundness": "3;3;4",
"soundness_avg": 3.3333333333333335,
"contribution": "2;2;3",
"contribution_avg": 2.3333333333333335,
"presentation": "4;3;4",
"presentation_avg": 3.6666666666666665
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:00.874573"
} | {
"id": "N7N7HNJnqr",
"metareview": "This paper considers gradient normalization and soft clipping methods in the context of Hamiltonian systems. Authors view these as a kind of Euler discretization of dissipative systems and use dynamical systems approach and show convergence to stationary points of the objective ... | {
"decision": "Reject"
} |
5wuZyG1ACs | 2409.15254v5 | Archon: An Architecture Search Framework for Inference-Time Techniques | {
"content": "## Abstract\n\nAbstract Inference-time techniques are emerging as highly effective tools to enhance large language model (LLM) capabilities. However, best practices for developing systems that combine these techniques remain underdeveloped due to our limited understanding of the utility of individual in... | [
{
"id": "Ie0q5x6E9x",
"initial_rating": 3,
"confidence": 3,
"soundness": 2,
"contribution": 2,
"presentation": 2,
"summary": "This paper presents ARCHON, a framework for stacking different LLM components to leverage test-time compute to improve performance. While a more general architect... | {
"rating": "3;3;6;6",
"rating_avg": 4.5,
"confidence": "4;3;2;3",
"confidence_avg": 3,
"soundness": "2;2;3;3",
"soundness_avg": 2.5,
"contribution": "2;2;2;3",
"contribution_avg": 2.25,
"presentation": "3;2;3;2",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:00.876328"
} | {
"id": "4h6lRsudDq",
"metareview": "This paper combines different LLMs and inference-time techniques to obtain better performance than with one of them alone. The problem of selecting the right combination of LLM and inference-time techniques is viewed as a hyperparameter optimization problem and addresses it with... | {
"decision": "Reject"
} |
5xbKFaaqkS | 2403.15740v2 | Protecting Copyrighted Material with Unique Identifiers in Large Language Model Training | {
"content": "## Abstract\n\nAbstract A major public concern regarding the training of large language models (LLMs) is whether they abusing copyrighted online text.\nPrevious membership inference methods may be misled by similar examples in vast amounts of training data.\nAdditionally, these methods are often too com... | [
{
"id": "jgzB3v4H8O",
"initial_rating": 5,
"confidence": 4,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "The paper proposes a method to detect the presence of copyrighted material of a user in the training corpus of LLMs. The authors suggest a method that uses unique... | {
"rating": "3;3;5;5;6",
"rating_avg": 4.4,
"confidence": "4;4;5;4;3",
"confidence_avg": 4,
"soundness": "3;3;2;3;3",
"soundness_avg": 2.8,
"contribution": "2;2;2;2;3",
"contribution_avg": 2.2,
"presentation": "2;3;2;3;3",
"presentation_avg": 2.6
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:00.877561"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
5xwx1Myosu | 2407.00957v2 | Expressivity of Neural Networks with Random Weights and Learned Biases | {
"content": "## Abstract\n\nAbstract Landmark universal function approximation results for neural networks with trained weights and biases provided impetus for the ubiquitous use of neural networks as learning models in Artificial Intelligence (AI) and neuroscience.\nRecent work has pushed the bounds of universal ap... | [
{
"id": "oiKQ0OIpjF",
"initial_rating": 5,
"confidence": 5,
"soundness": 3,
"contribution": 2,
"presentation": 4,
"summary": "Previous work has investigated the expressivity of feed-forward neural networks (FNNs) when only subsets of parameters are trained (ie. only the output layer, nor... | {
"rating": "5;5;5;8",
"rating_avg": 5.75,
"confidence": "4;4;5;3",
"confidence_avg": 4,
"soundness": "3;3;3;4",
"soundness_avg": 3.25,
"contribution": "3;2;2;3",
"contribution_avg": 2.5,
"presentation": "4;3;4;4",
"presentation_avg": 3.75
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:00.878409"
} | {
"id": "glvKErR2pL",
"metareview": "This paper makes a significant theoretical contribution by demonstrating that neural networks with fixed random weights and trainable biases are universal function approximators. The authors provide rigorous proofs complemented by empirical experiments, including tasks relevant... | {
"decision": "Accept (Poster)"
} |
5xxGP9x5dZ | 2410.09591v1 | Unlearn and Burn: Adversarial Machine Unlearning Requests Destroy Model Accuracy | {
"content": "## Abstract\n\nAbstract Machine unlearning algorithms, designed for selective removal of training data from models, have emerged as a promising approach to growing privacy concerns. In this work, we expose a critical yet underexplored vulnerability in the deployment of unlearning systems: the assumption... | [
{
"id": "VMpHssAE3Z",
"initial_rating": 8,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 4,
"summary": "This paper explores security risks within machine unlearning schemes, a topic gaining importance as the demand for ensuring \"the right to be forgotten\" grows. S... | {
"rating": "1;5;8",
"rating_avg": 4.666666666666667,
"confidence": "5;4;3",
"confidence_avg": 4,
"soundness": "3;2;3",
"soundness_avg": 2.6666666666666665,
"contribution": "1;2;3",
"contribution_avg": 2,
"presentation": "3;3;4",
"presentation_avg": 3.3333333333333335
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:00.879156"
} | {
"id": "V3JXfzymKn",
"metareview": "This paper received ratings of 8,8,3. The paper presents promising results on black-box attacks and machine unlearning but has several key weaknesses. It overlooks scalability challenges for high-dimensional data in black-box attacks, and fails to fully explore adaptive defense ... | {
"decision": "Accept (Poster)"
} |
5y3QbuK6HD | 2410.10578v3 | Burning RED: Unlocking Subtask-Driven Reinforcement Learning and Risk-Awareness in Average-Reward Markov Decision Processes | {
"content": "## Abstract\n\nAbstract Average-reward Markov decision processes (MDPs) provide a foundational framework for sequential decision-making under uncertainty. However, average-reward MDPs have remained largely unexplored in reinforcement learning (RL) settings, with the majority of RL-based efforts having b... | [
{
"id": "gTRsKqJl1A",
"initial_rating": 6,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "This paper studies a class of average-reward reinforcement learning problems, which includes risk-sensitive RL as special case. The main contribution is a new fr... | {
"rating": "3;3;3;6",
"rating_avg": 3.75,
"confidence": "3;4;3;3",
"confidence_avg": 3.25,
"soundness": "1;2;2;3",
"soundness_avg": 2,
"contribution": "3;2;2;3",
"contribution_avg": 2.5,
"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": "2026-01-14T22:16:00.880134"
} | {
"id": "8Fij7dH1Uv",
"metareview": "This paper introduces a Reward Extended Differential (RED) approach for risk-averse AMDP that aims to handle multiple subtasks concurrently, by defining the TD error through a modified reward generated by the observed rewards and subtasks through an invertible function. They lev... | {
"decision": "Reject"
} |
5z9GjHgerY | 2410.13782v1 | DPLM-2: A Multimodal Diffusion Protein Language Model | {
"content": "## Abstract\n\nAbstract Proteins are essential macromolecules defined by their amino acid sequences, which determine their three-dimensional structures and, consequently, their functions in all living organisms. Therefore, generative protein modeling necessitates a multimodal approach to simultaneously ... | [
{
"id": "iR1XwUrD8b",
"initial_rating": 3,
"confidence": 5,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "This paper introduces DPLM-2, a multimodal discrete diffusion protein language model that can simultaneously generate both protein sequences and their 3D structur... | {
"rating": "3;3;6",
"rating_avg": 4,
"confidence": "3;5;5",
"confidence_avg": 4.333333333333333,
"soundness": "2;3;3",
"soundness_avg": 2.6666666666666665,
"contribution": "3;2;3",
"contribution_avg": 2.6666666666666665,
"presentation": "2;3;4",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:00.881143"
} | {
"id": "P9ByjPrwdK",
"metareview": "This paper proposes a joint generative model for protein structure (represented by discrete tokens) and sequence. The paper details a comprehensive study. The authors and two of three reviewers engaged in a thorough discussion. Two referees are strongly supporting acceptance and... | {
"decision": "Accept (Poster)"
} |
5zGuFj0y9V | 2402.11566v2 | Boosting Semi-Supervised 2D Human Pose Estimation by Revisiting Data Augmentation and Consistency Training | {
"content": "## Abstract\n\nAbstract The 2D human pose estimation (HPE) is a basic visual problem. However, its supervised learning requires massive keypoint labels, which is labor-intensive to collect. Thus, we aim at boosting a pose estimator by excavating extra unlabeled data with semi-supervised learning (SSL). ... | [
{
"id": "5N9er3kUfb",
"initial_rating": 5,
"confidence": 5,
"soundness": 2,
"contribution": 2,
"presentation": 3,
"summary": "This paper presents a method for SSL 2D Human Pose estimation that improves existing methods from two aspects: data augmentation and better consistency loss desig... | {
"rating": "3;5;5;5;5",
"rating_avg": 4.6,
"confidence": "4;5;4;4;5",
"confidence_avg": 4.4,
"soundness": "2;2;3;2;2",
"soundness_avg": 2.2,
"contribution": "1;2;2;2;2",
"contribution_avg": 1.8,
"presentation": "2;2;3;3;3",
"presentation_avg": 2.6
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:00.881888"
} | {
"id": "OhkjSm1l3E",
"metareview": "This paper received five reviews, and the authors submitted a response addressing the queries raised. While two reviewer shifted to a positive stance after the rebuttal, albeit with some reservations, the others retained their initial ratings, which were largely below the accept... | {
"decision": "Reject"
} |
5zjsZiYEnr | 2411.06176v1 | M-Longdoc: A Benchmark For Multimodal Super-Long Document Understanding And A Retrieval-Aware Tuning Framework | {
"content": "## Abstract\n\nAbstract The ability to understand and answer questions over documents can be useful in many business and practical applications. However, documents often contain lengthy and diverse multimodal contents such as texts, figures, and tables, which are very time-consuming for humans to read t... | [
{
"id": "k3eDS7Fps6",
"initial_rating": 5,
"confidence": 4,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "This paper introduces a dataset for long document understanding challenges with an automatic evaluation approach. The authors also propose a retrieval-aware tunin... | {
"rating": "3;5;8",
"rating_avg": 5.333333333333333,
"confidence": "4;4;4",
"confidence_avg": 4,
"soundness": "2;3;3",
"soundness_avg": 2.6666666666666665,
"contribution": "2;2;3",
"contribution_avg": 2.3333333333333335,
"presentation": "2;3;3",
"presentation_avg": 2.6666666666666665
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:00.882599"
} | {
"id": "RE0bsrbn5p",
"metareview": "**Summary:**\nThe paper introduces M-LongDoc, a benchmark for evaluating multimodal models on lengthy documents containing text, figures, and tables, with open-ended questions requiring deep analysis. It proposes a retrieval-aware tuning framework to improve model performance by... | {
"decision": "Reject"
} |
60GeEoG5kD | 2410.02226v1 | Doubly Optimal Policy Evaluation for Reinforcement Learning | {
"content": "## Abstract\n\nAbstract Policy evaluation estimates the performance of a policy by (1) collecting data from the environment and (2) processing raw data into a meaningful estimate. Due to the sequential nature of reinforcement learning, any improper data-collecting policy or data-processing method substa... | [
{
"id": "E121iOJ6Fm",
"initial_rating": 6,
"confidence": 4,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "This paper studies the policy evaluation problem. Given a policy $\\pi$ to evaluate, the authors adopted the classic importance-weighting estimator with baseline ... | {
"rating": "3;3;8",
"rating_avg": 4.666666666666667,
"confidence": "3;4;4",
"confidence_avg": 3.6666666666666665,
"soundness": "2;1;3",
"soundness_avg": 2,
"contribution": "3;1;3",
"contribution_avg": 2.3333333333333335,
"presentation": "3;2;3",
"presentation_avg": 2.6666666666666665
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:00.883822"
} | {
"id": "gJwIx16Sl0",
"metareview": "In this paper, the authors propose a method to find a behavioral policy that can be used to estimate the value of a given target policy for a confidence level with small amount of data generation (policy rollouts). By optimizing the variance of the classic importance-weighting e... | {
"decision": "Accept (Poster)"
} |
60TXv9Xif5 | 2410.19746v1 | Metamizer: A Versatile Neural Optimizer for Fast and Accurate Physics Simulations | {
"content": "## Abstract\n\nAbstract Efficient physics simulations are essential for numerous applications, ranging from realistic cloth animations or smoke effects in video games, to analyzing pollutant dispersion in environmental sciences, to calculating vehicle drag coefficients in engineering applications. Unfor... | [
{
"id": "LuwxKfZu0i",
"initial_rating": 3,
"confidence": 4,
"soundness": 2,
"contribution": 3,
"presentation": 2,
"summary": "The paper proposes a meta-optimization technique for solving physics simulation problems, including linear and nonlinear PDE problems. The proposed Metamizer fra... | {
"rating": "3;3;3;6",
"rating_avg": 3.75,
"confidence": "4;4;4;4",
"confidence_avg": 4,
"soundness": "2;1;2;3",
"soundness_avg": 2,
"contribution": "2;2;3;3",
"contribution_avg": 2.5,
"presentation": "3;2;2;3",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:00.884692"
} | {
"id": "1KggmiP7hg",
"metareview": "**Summary and strengths** The paper presents a learned solver for partial-differential equations (PDEs) which minimizes a physics-based loss function. The method achieves high accuracy on a wide range of PDEs, such as Laplace, advection-diffusion and incompressible Navier-Stokes... | {
"decision": "Accept (Poster)"
} |
60i0ksMAhd | 2410.11689v1 | BlendRL: A Framework for Merging Symbolic and Neural Policy Learning | {
"content": "## Abstract\n\nAbstract Humans can leverage both symbolic reasoning and intuitive reactions. In contrast, reinforcement learning policies are typically encoded in either opaque systems like neural networks or symbolic systems that rely on predefined symbols and rules. This disjointed approach severely l... | [
{
"id": "z2N1mIv3AX",
"initial_rating": 5,
"confidence": 4,
"soundness": 1,
"contribution": 1,
"presentation": 2,
"summary": "This paper integrates condition-based logic decisions with neural network-based reinforcement learning policies through an LLM-based hybrid module to address the ... | {
"rating": "3;5;6;6",
"rating_avg": 5,
"confidence": "3;4;3;4",
"confidence_avg": 3.5,
"soundness": "1;3;3;2",
"soundness_avg": 2.25,
"contribution": "1;2;3;3",
"contribution_avg": 2.25,
"presentation": "2;2;4;3",
"presentation_avg": 2.75
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Spotlight",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:00.885500"
} | {
"id": "0Q6yTYP85s",
"metareview": "This paper presents BlendRL, a framework that integrates neural and symbolic policies for reinforcement learning, demonstrating improved performance over both pure neural networks and symbolic baseline approaches. The majority of reviewers praised the paper's clear presentation,... | {
"decision": "Accept (Spotlight)"
} |
61ss5RA1MM | 2410.18070v1 | Training Free Guided Flow-Matching with Optimal Control | {
"content": "## Abstract\n\nAbstract Controlled generation with pre-trained Diffusion and Flow Matching models has vast applications. One strategy for guiding ODE-based generative models is through optimizing a target loss R ( x 1 ) 𝑅 subscript 𝑥 1 R(x_{1}) italic_R ( italic_x start_POSTSUBSCRIPT 1 end_POSTSUBSC... | [
{
"id": "haMe8Qw8Jm",
"initial_rating": 6,
"confidence": 3,
"soundness": 2,
"contribution": 2,
"presentation": 3,
"summary": "This paper proposed a new framework for controlled generation using pre-trained diffusion and flow matching models, dubbed OC-Flow. The method is based on sound t... | {
"rating": "3;5;6;6",
"rating_avg": 5,
"confidence": "3;3;3;2",
"confidence_avg": 2.75,
"soundness": "1;2;3;4",
"soundness_avg": 2.5,
"contribution": "2;2;2;4",
"contribution_avg": 2.5,
"presentation": "3;1;3;3",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:00.886299"
} | {
"id": "Wuho9Fl9WN",
"metareview": "Among other things, the reviewers have highlighted the nice framing of guided flow matching as an optimal control problem, the contribution to formalizing and extending the existing guided-flow matching techniques to SO(3), and the applicability of the proposed approach to vario... | {
"decision": "Accept (Poster)"
} |
634kHJgaOL | 2405.20179v2 | ROBO-INSTRUCT: Simulator-Augmented Instruction Alignment For Finetuning Code LLMs | {
"content": "## Abstract\n\nAbstract Open-weight LLMs are particularly appealing choices to generate training data for fine-tuning Code LLMs on domain-specific service robot applications because they are cost-effective, customizable, and offer better privacy protection. However, unlike proprietary LLMs, open-weight ... | [
{
"id": "TZC3DCZzCO",
"initial_rating": 6,
"confidence": 2,
"soundness": 4,
"contribution": 2,
"presentation": 4,
"summary": "The goal of this paper is to",
"strengths": "Clarity. The authors did a phenomenal job describing their method and experimental process with precision. The sp... | {
"rating": "3;5;5;6",
"rating_avg": 4.75,
"confidence": "3;4;4;2",
"confidence_avg": 3.25,
"soundness": "1;2;3;4",
"soundness_avg": 2.5,
"contribution": "2;3;2;2",
"contribution_avg": 2.25,
"presentation": "2;3;3;4",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:00.887156"
} | {
"id": "1F2SZreZgR",
"metareview": "This paper proposes a framework to fine-tune a code LLM for robotics tasks. The paper is well written and motivated. However, at the same time the reviewers have raised several concerns, particularly on the novelty of ideas in the paper. The evaluated tasks and the real world de... | {
"decision": "Reject"
} |
63r2sTjkCv | 2410.08938v1 | KinDEL: DNA-Encoded Library Dataset for Kinase Inhibitors | {
"content": "## Abstract\n\nAbstract DNA-Encoded Libraries (DEL) are combinatorial small molecule libraries that offer an efficient way to characterize diverse chemical spaces. Selection experiments using DELs are pivotal to drug discovery efforts, enabling high-throughput screens for hit finding. However, limited a... | [
{
"id": "8yMp6iWN9I",
"initial_rating": 5,
"confidence": 3,
"soundness": 2,
"contribution": 3,
"presentation": 2,
"summary": "The authors have released a new dataset, KinDEL, based on DNA-encoded library (DEL) testing, specifically targeting two kinases, MAPK14 and DDR1. They conducted e... | {
"rating": "5;5;5;6",
"rating_avg": 5.25,
"confidence": "3;3;3;4",
"confidence_avg": 3.25,
"soundness": "3;3;2;4",
"soundness_avg": 3,
"contribution": "2;2;3;3",
"contribution_avg": 2.5,
"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": "2026-01-14T22:16:00.888881"
} | {
"id": "wC5tUgsyyF",
"metareview": "This work aims to present a large publicly accessible DNA-encoded library (DEL) datasets, referred to as KinDEL, which comprises two kinases: namely, Mitogen-Activated Protein Kinase 14 (MAPK14) and Discoidin Domain Receptor Tyrosine Kinase 1 (DDR1).\nAll reviewers recognize the... | {
"decision": "Reject"
} |
66NzcRQuOq | 2410.05954v1 | Pyramidal Flow Matching for Efficient Video Generative Modeling | {
"content": "## Abstract\n\nAbstract Video generation requires modeling a vast spatiotemporal space, which demands significant computational resources and data usage. To reduce the complexity, the prevailing approaches employ a cascaded architecture to avoid direct training with full resolution. Despite reducing com... | [
{
"id": "0teSp6SxL8",
"initial_rating": 5,
"confidence": 4,
"soundness": 2,
"contribution": 3,
"presentation": 2,
"summary": "This paper introduces a novel pyramidal flow matching scheme for video generation, which significantly improves training efficiency while preserving generation qu... | {
"rating": "5;5;6;8;8",
"rating_avg": 6.4,
"confidence": "4;4;5;4;5",
"confidence_avg": 4.4,
"soundness": "2;2;2;4;4",
"soundness_avg": 2.8,
"contribution": "2;3;3;4;4",
"contribution_avg": 3.2,
"presentation": "3;2;3;2;4",
"presentation_avg": 2.8
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:00.889588"
} | {
"id": "N1xNwjMdvU",
"metareview": "The paper presents an important step towards efficient training of text-to-video generation. Such models are known to be very hard to train, requiring a substantial computational budget. Here, we are presented with a method to reinterpret denoising trajectory as a series of pyra... | {
"decision": "Accept (Poster)"
} |
67X93aZHII | 2410.19735v1 | Model merging with SVD to tie the Knots | {
"content": "## Abstract\n\nAbstract Recent model merging methods demonstrate that the parameters of fully-finetuned models specializing in distinct tasks can be combined into one model capable of solving all tasks without retraining. Yet, this success does not transfer well when merging LoRA finetuned models. We st... | [
{
"id": "86T8BfNnts",
"initial_rating": 8,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "The paper titled \"Model Merging with SVD to Tie the Knots\" explores the challenge of merging Low-Rank Adaptation (LoRA) finetuned models. While model merging ha... | {
"rating": "3;6;6;8",
"rating_avg": 5.75,
"confidence": "3;4;3;3",
"confidence_avg": 3.25,
"soundness": "2;3;3;3",
"soundness_avg": 2.75,
"contribution": "2;2;3;3",
"contribution_avg": 2.5,
"presentation": "2;3;3;3",
"presentation_avg": 2.75
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:00.890343"
} | {
"id": "Hfw6sfNd66",
"metareview": "This paper proposes a method for model-merging that is designed for LoRA-finetuned models. Although methods such as TIES and DARE are well known for model-merging, they do not perform well for LoRA-trained models. The authors show that this is because LoRA-parameters are not wel... | {
"decision": "Accept (Poster)"
} |
6AUzsrsNUx | 2407.12871v2 | MetaTool: Facilitating Large Language Models to Master Tools with Meta-task Augmentation | {
"content": "## Abstract\n\nAbstract Utilizing tools with Large Language Models (LLMs) is essential for grounding AI agents in real-world applications. The prevailing approach involves few-shot prompting with demonstrations or fine-tuning with expert annotations. However, mere in-context demonstrations may fail to c... | [
{
"id": "RqU8G0OFcJ",
"initial_rating": 3,
"confidence": 3,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "This paper proposes to achieve generalizable tool learning by additionally training models on meta-reasoning QA tasks. The meta-reasoning data are constructed by ... | {
"rating": "3;5;5;6",
"rating_avg": 4.75,
"confidence": "3;3;4;4",
"confidence_avg": 3.5,
"soundness": "2;3;2;3",
"soundness_avg": 2.5,
"contribution": "2;3;3;2",
"contribution_avg": 2.5,
"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": "2026-01-14T22:16:00.890942"
} | {
"id": "Izlh35BXBb",
"metareview": "This paper introduces MetaTool, a novel approach to enhance large language models’ (LLMs) ability to use tools. Unlike traditional methods that rely on prompts or labeled data, MetaTool employs self-supervised learning through six meta-tasks: Effect, Decision-making, Reversion, ... | {
"decision": "Reject"
} |
6Ai8SuDsh3 | 2410.15910v2 | Diverse Policies Recovering via Pointwise Mutual Information Weighted Imitation Learning | {
"content": "## Abstract\n\nAbstract Recovering a spectrum of diverse policies from a set of expert trajectories is an important research topic in imitation learning. After determining a latent style for a trajectory, previous diverse policies recovering methods usually employ a vanilla behavioral cloning learning o... | [
{
"id": "bmlIex7WtW",
"initial_rating": 5,
"confidence": 4,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "The paper explores a methodology for deriving diverse policies from expert trajectory data. Rising from the traditional conditional behavior cloning (BC) algorith... | {
"rating": "5;6;6;8",
"rating_avg": 6.25,
"confidence": "4;4;2;4",
"confidence_avg": 3.5,
"soundness": "3;3;2;3",
"soundness_avg": 2.75,
"contribution": "2;3;3;4",
"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": "2026-01-14T22:16:00.891525"
} | {
"id": "macBImsvgM",
"metareview": "Summary: This paper proposes Behavioral Cloning with Pointwise Mutual Information Weighting (BC-PMI), a method that enhances policy diversity by weighting state-action pairs based on their relevance to trajectory styles using Pointwise Mutual Information (PMI). Experimental resu... | {
"decision": "Accept (Poster)"
} |
6ApaDkSMtX | 2410.01600v1 | Encoder-only Next Token Prediction | {
"content": "## Abstract\n\nAbstract Next-token prediction models have predominantly relied on decoder-only Transformers with causal attention, driven by the common belief that causal attention is essential to prevent “cheating” by masking future tokens. We challenge this widely accepted notion and argue that this d... | [
{
"id": "gruzGltLdZ",
"initial_rating": 5,
"confidence": 3,
"soundness": 3,
"contribution": 2,
"presentation": 2,
"summary": "This paper analyzes the Encoder-only Next Token Prediction (ENTP), which challenges the prevailing use of decoder-only Transformers with causal attention in next-... | {
"rating": "3;5;5;6",
"rating_avg": 4.75,
"confidence": "3;4;3;4",
"confidence_avg": 3.5,
"soundness": "3;2;3;4",
"soundness_avg": 3,
"contribution": "1;2;2;2",
"contribution_avg": 1.75,
"presentation": "1;2;2;4",
"presentation_avg": 2.25
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:00.892157"
} | {
"id": "wDgXqeTXce",
"metareview": "The premise of this paper is straightforward -- the Authors argue for a critical look at using encoder-only (\"BERT-style\") Transformers for next-token prediction training. They provide evidence for this claim through three avenues: (a) theoretical analysis, (b) synthetic count... | {
"decision": "Reject"
} |
6BjEqGn1OO | 2405.13203v1 | Modeling Real-Time Interactive Conversations as Timed Diarized Transcripts | {
"content": "## Abstract\n\nAbstract Chatbots built upon language models have exploded in popularity, but they have largely been limited to synchronous, turn-by-turn dialogues. In this paper we present a simple yet general method to simulate real-time interactive conversations using pretrained text-only language mod... | [
{
"id": "117o7KNRkS",
"initial_rating": 8,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "The paper describes a method for applying a standard LLM to the task of generating dialogues incrementally, either turn-by-turn or word-by-word, by considering an... | {
"rating": "3;5;5;8",
"rating_avg": 5.25,
"confidence": "2;4;2;3",
"confidence_avg": 2.75,
"soundness": "2;2;3;3",
"soundness_avg": 2.5,
"contribution": "2;2;3;3",
"contribution_avg": 2.5,
"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": "2026-01-14T22:16:00.893166"
} | {
"id": "ou7STJQKqs",
"metareview": "This paper presents an interesting method for creating transcripts, however reviewers felt that the work was not well positioned with respect to other existing work in the area (references have been given by individual reviewers) and that the method was only tested on very narro... | {
"decision": "Reject"
} |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.