paper_id string | arxiv_id string | title string | markdown dict | reviews list | scores dict | metadata dict | meta_review dict | decision dict |
|---|---|---|---|---|---|---|---|---|
KI1zldOFz9 | 2406.10126v2 | Training-free Camera Control for Video Generation | {
"content": "## Abstract\n\nAbstract We propose a training-free and robust solution to offer camera movement control for off-the-shelf video diffusion models. Unlike previous work, our method does not require any supervised finetuning\non camera-annotated datasets or self-supervised training via data augmentation. I... | [
{
"id": "5Dmx6irYDe",
"initial_rating": 5,
"confidence": 5,
"soundness": 2,
"contribution": 3,
"presentation": 3,
"summary": "The paper introduces a method named CamTrol to enable controllable camera movements for video generation using off-the-shelf video diffusion models. The innovatio... | {
"rating": "3;3;5;5;6",
"rating_avg": 4.4,
"confidence": "5;5;4;5;4",
"confidence_avg": 4.6,
"soundness": "3;2;2;2;4",
"soundness_avg": 2.6,
"contribution": "2;2;2;3;3",
"contribution_avg": 2.4,
"presentation": "2;2;3;3;3",
"presentation_avg": 2.6
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.737883"
} | {
"id": "cNcGtx9GiL",
"metareview": "The paper uses pre-trained video diffusion models to introduce a training-free method for controllable camera movements in video generation. By leveraging 3D point cloud modeling, camera motion simulation, and layout priors in noisy latent, it achieves flexible camera control wi... | {
"decision": "Accept (Poster)"
} |
KIgaAqEFHW | 2408.03350v2 | miniCTX: Neural Theorem Proving with (Long-)Contexts | {
"content": "## Abstract\n\nAbstract Real-world formal theorem proving often depends on a wealth of context, including definitions, lemmas, comments, file structure, and other information. We introduce miniCTX , which tests a model’s ability to prove formal mathematical theorems that depend on new context that is no... | [
{
"id": "bHfIs5AN3M",
"initial_rating": 8,
"confidence": 3,
"soundness": 3,
"contribution": 4,
"presentation": 4,
"summary": "This paper introduces $\\texttt{miniCTX}$, a benchmark of 384 problems with context-rich information, e.g., in-file definitions and lemmas, and evaluates the mode... | {
"rating": "5;6;8",
"rating_avg": 6.333333333333333,
"confidence": "4;3;3",
"confidence_avg": 3.3333333333333335,
"soundness": "2;3;3",
"soundness_avg": 2.6666666666666665,
"contribution": "3;3;4",
"contribution_avg": 3.3333333333333335,
"presentation": "2;3;4",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Oral",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.739989"
} | {
"id": "GGTdsqFD7H",
"metareview": "This paper introduces miniCTX, a novel benchmark for evaluating neural theorem-proving models in real-world, context-rich scenarios, alongside NTP-toolkit, an automated tool for data extraction and annotation. The propositions are well-motivated, addressing the limitations of ex... | {
"decision": "Accept (Oral)"
} |
KJzz4UwqTb | 2402.04902v4 | L4Q: Parameter Efficient Quantization-Aware Fine-Tuning on Large Language Models | {
"content": "## Abstract\n\nAbstract Due to the high memory and computational costs associated with large language models (LLMs), model compression techniques such as quantization, which reduces inference costs, and parameter-efficient fine-tuning (PEFT) methods like Low-Rank Adaptation (LoRA), which reduce training... | [
{
"id": "beIzc7gQzJ",
"initial_rating": 5,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "The paper aims to reduce the costs of LLMs by compressing them, specifically, discusses the quantization-aware training technique for reducing inference costs wit... | {
"rating": "3;5;5;5",
"rating_avg": 4.5,
"confidence": "4;4;4;4",
"confidence_avg": 4,
"soundness": "1;3;3;3",
"soundness_avg": 2.5,
"contribution": "2;2;2;3",
"contribution_avg": 2.25,
"presentation": "1;3;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": "2026-01-14T22:16:01.743197"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
KL8Sm4xRn7 | 2410.09230v2 | Improving Semantic Understanding in Speech Language Models via Brain-tuning | {
"content": "## Abstract\n\nAbstract Speech language models align with human brain responses to natural language to an impressive degree. However, current models rely heavily on low-level speech features, indicating they lack brain-relevant semantics which limits their utility as model organisms of semantic processi... | [
{
"id": "i9b3hlueJx",
"initial_rating": 5,
"confidence": 3,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "- In short the main idea is to fine-tune existing speech models so they are better aligned with fMRI recordings and then show that this results in better, more se... | {
"rating": "5;5;6;6",
"rating_avg": 5.5,
"confidence": "4;3;4;3",
"confidence_avg": 3.5,
"soundness": "3;3;3;4",
"soundness_avg": 3.25,
"contribution": "3;2;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": "2026-01-14T22:16:01.745592"
} | {
"id": "QvTUWuV2v4",
"metareview": "The paper introduces \"brain-tuning,\" a method for fine-tuning speech language models (SLMs) using fMRI data from individuals listening to natural stories. The central claim is that incorporating brain signals into the training process enhances the models' semantic understandin... | {
"decision": "Accept (Poster)"
} |
KLIN1QdcX4 | 2406.19668v1 | PopAlign: Population-Level Alignment for Fair Text-to-Image Generation | {
"content": "## Abstract\n\nAbstract Text-to-image (T2I) models achieve high-fidelity generation through extensive training on large datasets. However, these models may unintentionally pick up undesirable biases of their training data, such as over-representation of particular identities in gender or ethnicity neutr... | [
{
"id": "2Ootsu0GFV",
"initial_rating": 3,
"confidence": 4,
"soundness": 3,
"contribution": 2,
"presentation": 2,
"summary": "This paper addresses the issue of inherent biases in generative models, which often stem from biased large-scale training data. The authors propose a method to re... | {
"rating": "3;3;5",
"rating_avg": 3.6666666666666665,
"confidence": "3;4;3",
"confidence_avg": 3.3333333333333335,
"soundness": "2;3;3",
"soundness_avg": 2.6666666666666665,
"contribution": "3;2;2",
"contribution_avg": 2.3333333333333335,
"presentation": "3;2;2",
"presentation_avg": 2.3333333333333... | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:01.746493"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
KMCJXjlDDr | 2410.04803v2 | Timer-XL: Long-Context Transformers for Unified Time Series Forecasting | {
"content": "## Abstract\n\nAbstract We present Timer-XL, a generative Transformer for unified time series forecasting. To uniformly predict 1D and 2D time series, we generalize next token prediction, predominantly adopted for causal generation of 1D sequences, to multivariate next token prediction . The proposed pa... | [
{
"id": "I1nTRDjveL",
"initial_rating": 5,
"confidence": 5,
"soundness": 2,
"contribution": 2,
"presentation": 2,
"summary": "The paper proposes Timer-XL, a transformer decoder model for time series forecasting. Building upon the existing Timer model, Timer-XL extends the model with a lo... | {
"rating": "3;5;6",
"rating_avg": 4.666666666666667,
"confidence": "4;5;4",
"confidence_avg": 4.333333333333333,
"soundness": "2;2;3",
"soundness_avg": 2.3333333333333335,
"contribution": "2;2;3",
"contribution_avg": 2.3333333333333335,
"presentation": "2;2;3",
"presentation_avg": 2.333333333333333... | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.747263"
} | {
"id": "6kgHckgHq5",
"metareview": "This paper introduces a generative transformer model for unified timeseries forecasting. They propose a new form of TimeAttention which allows for the application of generative transformers on multivariate data and demonstrate improvements in long-context regimes. They provide r... | {
"decision": "Accept (Poster)"
} |
KMWGzQi7Qy | 2406.07780v1 | A Critical Look At Tokenwise Reward-Guided Text Generation | {
"content": "## Abstract\n\nAbstract Large language models (LLMs) can significantly be improved by aligning to human preferences—the so-called reinforcement learning from human feedback (RLHF). However, the cost of fine-tuning an LLM is prohibitive for many users.\nDue to their ability to bypass LLM finetuning, toke... | [
{
"id": "TZ3XPmUfU4",
"initial_rating": 3,
"confidence": 3,
"soundness": 3,
"contribution": 1,
"presentation": 3,
"summary": "This paper pointed out that current RGTG methods struggle as they rely on full-sequence reward models for partial sequences. To improve this, the authors augment ... | {
"rating": "3;5;6;8",
"rating_avg": 5.5,
"confidence": "3;4;4;3",
"confidence_avg": 3.5,
"soundness": "3;2;3;3",
"soundness_avg": 2.75,
"contribution": "1;2;3;3",
"contribution_avg": 2.25,
"presentation": "3;2;4;3",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:01.747991"
} | {
"id": "6pmZSWEGXh",
"metareview": "#### **Summary** \nThe paper proposes a new method PARGS to address the limitations of reward-guided text generation (RGTG) by training a Bradley-Terry (BT) reward model on partial sequences instead of full sequences. This theoretically principled method provides a clear connec... | {
"decision": "Reject"
} |
KQsR2JrCwx | 2409.20434v1 | QAEncoder: Towards Aligned Representation Learning in Question Answering System | {
"content": "## Abstract\n\nAbstract. Modern QA systems entail retrieval-augmented generation (RAG) for accurate and trustworthy responses. However, the inherent gap between user queries and relevant documents hinders precise matching. Motivated by our conical distribution hypothesis, which posits that potential que... | [
{
"id": "TvKIqpoJLq",
"initial_rating": 6,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "The paper proposes a simple but novel document encoding method based on a set of these corresponding (automatically generated) queries, QAEncoder, which generates... | {
"rating": "3;5;5;6",
"rating_avg": 4.75,
"confidence": "3;5;4;4",
"confidence_avg": 4,
"soundness": "2;2;3;3",
"soundness_avg": 2.5,
"contribution": "2;2;2;3",
"contribution_avg": 2.25,
"presentation": "2;3;3;3",
"presentation_avg": 2.75
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:01.748715"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
KRnsX5Em3W | 2410.02707v3 | LLMs Know More Than They Show: On the Intrinsic Representation of LLM Hallucinations | {
"content": "## Abstract\n\nAbstract Large language models (LLMs) often produce errors, including factual inaccuracies, biases, and reasoning failures, collectively referred to as “hallucinations”.\nRecent studies have demonstrated that LLMs’ internal states encode information regarding the truthfulness of their out... | [
{
"id": "JpPatFRyWw",
"initial_rating": 6,
"confidence": 3,
"soundness": 2,
"contribution": 2,
"presentation": 3,
"summary": "This paper examines the internal representations of large language models (LLMs) to understand how they encode truthfulness and produce hallucinations—errors such... | {
"rating": "5;6;6;6",
"rating_avg": 5.75,
"confidence": "4;4;3;3",
"confidence_avg": 3.5,
"soundness": "4;3;3;2",
"soundness_avg": 3,
"contribution": "3;2;3;2",
"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": "2026-01-14T22:16:01.749697"
} | {
"id": "EpHXfWnEQY",
"metareview": "This paper proposes a new approach to investigate hallucinations in LLMs by using probing classifiers to analyze their internal representations. Reviewers found that the paper provides some interesting findings, e.g., the internal representations of LLMs' responses are skill spe... | {
"decision": "Accept (Poster)"
} |
KRv9NubipP | 2411.04679v1 | CaPo: Cooperative Plan Optimization for Efficient Embodied Multi-Agent Cooperation | {
"content": "## Abstract\n\nAbstract In this work, we address the cooperation problem among large language model (LLM) based embodied agents, where agents must cooperate to achieve a common goal.\nPrevious methods often execute actions extemporaneously and incoherently, without long-term strategic and cooperative pl... | [
{
"id": "gSZpgcsb2a",
"initial_rating": 6,
"confidence": 4,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "The paper introduced Cooperative Plan Optimization into the CoELA framework to improve embodied multi-agent cooperation. A new meta-plan mechanism ensures agents ... | {
"rating": "5;5;6;6",
"rating_avg": 5.5,
"confidence": "3;5;4;4",
"confidence_avg": 4,
"soundness": "3;2;4;2",
"soundness_avg": 2.75,
"contribution": "2;1;3;2",
"contribution_avg": 2,
"presentation": "3;3;4;3",
"presentation_avg": 3.25
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.750754"
} | {
"id": "61fc2WCsvc",
"metareview": "This paper proposes an LLM-based embodied cooperation method that extends prior methods such as CoELA with a meta plan generation and dynamic adjustment, inspired by human cooperation. This is a meaningful contribution to LLM-based multi-agent cooperation. The revision and addit... | {
"decision": "Accept (Poster)"
} |
KSLkFYHlYg | 2411.04130v1 | ShEPhERD: Diffusing shape, electrostatics, and pharmacophores for bioisosteric drug design | {
"content": "## Abstract\n\nAbstract Engineering molecules to exhibit precise 3D intermolecular interactions with their environment forms the basis of chemical design.\nIn ligand-based drug design, bioisosteric analogues of known bioactive hits are often identified by virtually screening chemical libraries with shap... | [
{
"id": "MW2n1oeP5f",
"initial_rating": 6,
"confidence": 5,
"soundness": 2,
"contribution": 4,
"presentation": 4,
"summary": "This paper introduces ShEPhERD, a generative SE(3)-equivariant diffusion model for molecular drug design that incorporates 3D structural features inspired by medi... | {
"rating": "6;6;10",
"rating_avg": 7.333333333333333,
"confidence": "4;5;5",
"confidence_avg": 4.666666666666667,
"soundness": "3;2;4",
"soundness_avg": 3,
"contribution": "3;4;4",
"contribution_avg": 3.6666666666666665,
"presentation": "3;4;4",
"presentation_avg": 3.6666666666666665
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Oral",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.751766"
} | {
"id": "vJ6lx2NFCU",
"metareview": "In this work, authors introduce ShEPhERD, an SE(3)-equivariant diffusion model for 3D molecular generation that jointly models molecular graphs, shapes, electrostatic surfaces, and pharmacophores. The model enables conditional generation of molecules with desired 3D interaction ... | {
"decision": "Accept (Oral)"
} |
KSPBh07jEO | 2411.06528v1 | Epistemic Integrity in Large Language Models | {
"content": "## Abstract\n\nAbstract Large language models are increasingly relied upon as sources of information, but their propensity for generating false or misleading statements with high confidence poses risks for users and society. In this paper, we confront the critical problem of epistemic miscalibration — w... | [
{
"id": "gDTQPUGKep",
"initial_rating": 3,
"confidence": 4,
"soundness": 1,
"contribution": 1,
"presentation": 2,
"summary": "The paper addresses an issue in the deployment of large language models (LLMs): epistemic miscalibration, where a model’s expressed confidence does not match its ... | {
"rating": "3;3;3;6",
"rating_avg": 3.75,
"confidence": "4;4;4;4",
"confidence_avg": 4,
"soundness": "2;1;1;3",
"soundness_avg": 1.75,
"contribution": "2;1;1;3",
"contribution_avg": 1.75,
"presentation": "2;1;2;3",
"presentation_avg": 2
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:01.753086"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
KW6B6s1X82 | 2410.18538v1 | SMITE: Segment Me In TimE | {
"content": "## Abstract\n\nAbstract Segmenting an object in a video presents significant challenges. Each pixel must be accurately labeled, and these labels must remain consistent across frames. The difficulty increases when the segmentation is with arbitrary granularity, meaning the number of segments can vary arb... | [
{
"id": "xdfIbElj3r",
"initial_rating": 6,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "This paper defines a task of flexible granularity, which aims to segment user-intended granularity temporally using one or a few reference images. To do so, it pr... | {
"rating": "5;6;6;6;8",
"rating_avg": 6.2,
"confidence": "3;4;3;4;3",
"confidence_avg": 3.4,
"soundness": "3;3;2;3;3",
"soundness_avg": 2.8,
"contribution": "2;2;3;3;3",
"contribution_avg": 2.6,
"presentation": "3;3;2;3;3",
"presentation_avg": 2.8
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.754093"
} | {
"id": "l83L4AROKF",
"metareview": "This paper addresses an interesting problem in the video-object segmentation (VOS) space. \n\nMost VOS methods assume an object query in the first video frame (queries can come in the form of segmentation masks, points, or bounding boxes), and the task is to segment the queried ... | {
"decision": "Accept (Poster)"
} |
KWo4w1UXs8 | 2409.11689v1 | GUNet: A Graph Convolutional Network United Diffusion Model for Stable and Diversity Pose Generation | {
"content": "## Abstract\n\nAbstract Pose skeleton images are an important reference in pose-controllable image generation. In order to enrich the source of skeleton images, recent works have investigated the generation of pose skeletons based on natural language. These methods are based on GANs. However, it remains... | [
{
"id": "OnL1oMsqOu",
"initial_rating": 3,
"confidence": 3,
"soundness": 2,
"contribution": 2,
"presentation": 2,
"summary": "The paper proposes a diffusion model-based framework for the task of generating diverse and skeletally structurally stable 2D human pose skeletons (text2pose). Th... | {
"rating": "3;3;3",
"rating_avg": 3,
"confidence": "4;4;3",
"confidence_avg": 3.6666666666666665,
"soundness": "1;2;2",
"soundness_avg": 1.6666666666666667,
"contribution": "1;2;2",
"contribution_avg": 1.6666666666666667,
"presentation": "3;1;2",
"presentation_avg": 2
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:01.754821"
} | {
"id": "qIaRL065Rl",
"metareview": "The paper addresses the task of generating diverse and structurally stable 2D human pose skeletons from text descriptions using graph convolutional networks to capture spatial relationships in human poses. The reviewers agree that capturing the relation between keypoints using a... | {
"decision": "Reject"
} |
KX5hd1RhYP | 2410.06895v1 | Average Certified Radius is a Poor Metric for Randomized Smoothing | {
"content": "## Abstract\n\nAbstract Randomized smoothing is a popular approach for providing certified robustness guarantees against adversarial attacks, and has become a very active area of research. Over the past years, the average certified radius (ACR) has emerged as the single most important metric for compari... | [
{
"id": "mwHHU35ap4",
"initial_rating": 5,
"confidence": 5,
"soundness": 3,
"contribution": 1,
"presentation": 3,
"summary": "This submission studies the metric of average certified radius (ACR) in randomized smoothing training literature. First, the submission theoretically shows that A... | {
"rating": "3;5;5",
"rating_avg": 4.333333333333333,
"confidence": "4;5;4",
"confidence_avg": 4.333333333333333,
"soundness": "3;3;3",
"soundness_avg": 3,
"contribution": "1;3;2",
"contribution_avg": 2,
"presentation": "3;2;3",
"presentation_avg": 2.6666666666666665
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:01.755547"
} | {
"id": "MeXcMMw6Jl",
"metareview": "This paper is the first to show that the average certified radius (ACR) is a poor metric for evaluating robustness provided by randomized smoothing. The idea is interesting. But reviewers think the technical contribution is limited and topics of this paper is too narrow. It is ... | {
"decision": "Reject"
} |
KXDOmD7DM7 | 2410.15346v1 | YOLO-RD: Introducing Relevant and Compact Explicit Knowledge to YOLO by Retriever-Dictionary | {
"content": "## Abstract\n\nAbstract Identifying and localizing objects within images is a fundamental challenge, and numerous efforts have been made to enhance model accuracy by experimenting with diverse architectures and refining training strategies. Nevertheless, a prevalent limitation in existing models is over... | [
{
"id": "EvLC4RVLYV",
"initial_rating": 3,
"confidence": 4,
"soundness": 2,
"contribution": 2,
"presentation": 3,
"summary": "This paper aims to improve object detection with YOLO by introducing explicit knowledge in the form of a Retriever-Dictionary (RD) module. The dictionary is creat... | {
"rating": "3;6;6;6",
"rating_avg": 5.25,
"confidence": "4;4;4;3",
"confidence_avg": 3.75,
"soundness": "2;3;3;3",
"soundness_avg": 2.75,
"contribution": "2;3;3;2",
"contribution_avg": 2.5,
"presentation": "3;2;4;3",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.756522"
} | {
"id": "HoUxLqTdLo",
"metareview": "This paper proposes a Retriever Dictionary (RD) module to introduce explicit knowledge of a dataset for enhancing object detection. Experiments with multiple detectors, including YOLOv7, YOLOv9, Faster R-CNN, and Deformable DETR, demonstrate performance improvements.\n\nThe main... | {
"decision": "Accept (Poster)"
} |
KYOdZRR6nr | 2410.10037v1 | GALA: Geometry-Aware Local Adaptive Grids for Detailed 3D Generation | {
"content": "## Abstract\n\nAbstract We propose GALA, a novel representation of 3D shapes that (i) excels at capturing and reproducing complex geometry and surface details, (ii) is computationally efficient, and (iii) lends itself to 3D generative modelling with modern, diffusion-based schemes. The key idea of GALA ... | [
{
"id": "21wbqvNQ7s",
"initial_rating": 8,
"confidence": 5,
"soundness": 3,
"contribution": 3,
"presentation": 4,
"summary": "The paper presents a storage-efficient new representation of 3D triangle surface meshes, GALA, which can be used for 3D generation trained by a transformer. To cr... | {
"rating": "5;6;8;8",
"rating_avg": 6.75,
"confidence": "4;4;2;5",
"confidence_avg": 3.75,
"soundness": "2;4;4;3",
"soundness_avg": 3.25,
"contribution": "3;3;3;3",
"contribution_avg": 3,
"presentation": "3;4;2;4",
"presentation_avg": 3.25
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.757569"
} | {
"id": "hc9XmPfX4s",
"metareview": "In this paper, the authors propose a novel representation of 3D shapes called GALA, which is able to represent complex geometric details with efficient memory cost. The method leverages the global sparsity of surfaces within a 3D volume and their local surface properties, and bu... | {
"decision": "Accept (Poster)"
} |
KYipmCMmSO | 2402.18905v1 | Characterizing the Training Dynamics of Private Fine-tuning with Langevin Diffusion | {
"content": "## Abstract\n\nAbstract Differentially private (DP) machine learning pipelines typically involve a two-phase process: non-private pre-training on a public dataset, followed by fine-tuning on private data using DP optimization techniques. In the DP setting, it has been observed that full fine-tuning may ... | [
{
"id": "qDxm4JMTCb",
"initial_rating": 5,
"confidence": 2,
"soundness": 3,
"contribution": 2,
"presentation": 2,
"summary": "This paper addresses the challenge of differentially private fine-tuning of deep learning models. The authors highlight that naïve full-parameter fine-tuning lead... | {
"rating": "5;6;8",
"rating_avg": 6.333333333333333,
"confidence": "2;3;2",
"confidence_avg": 2.3333333333333335,
"soundness": "3;3;3",
"soundness_avg": 3,
"contribution": "2;3;3",
"contribution_avg": 2.6666666666666665,
"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": "2026-01-14T22:16:01.759732"
} | {
"id": "lXQeRO7tJx",
"metareview": "This paper is in borderline! Although the authors have addressed some comments and questions, the reviewer is still not convinced about the theoretical contributions of the paper as they cannot see the additional value to experimental results. Moreover, the reviewer also thinks ... | {
"decision": "Reject"
} |
KZgo2YQbhc | 2406.05641v1 | PaRa: Personalizing Text-to-Image Diffusion via Parameter Rank Reduction | {
"content": "## Appendix A Theorem and Proof of PRR\n\n###### Theorem 1 .\n\nFor matrix $W$, the image space of $W$ is a d-dimension vector space $S_{d}$. If we have matrix $Q=\\begin{bmatrix}\\vec{q_{1}}\\ \\vec{q_{2}}\\ ...\\ \\vec{q_{r}}\\end{bmatrix}$ with $q_{i}\\in S_{d}$ and vectors $q_{i}$ are mutually ortho... | [
{
"id": "FaotQdOFxC",
"initial_rating": 6,
"confidence": 4,
"soundness": 3,
"contribution": 4,
"presentation": 3,
"summary": "This paper proposes a framework for T2I model personalization called PaRa (Parameter Rank Reduction). It introduces an innovative approach to control the rank of ... | {
"rating": "5;6;8;8",
"rating_avg": 6.75,
"confidence": "3;4;3;4",
"confidence_avg": 3.5,
"soundness": "3;3;3;3",
"soundness_avg": 3,
"contribution": "3;4;3;3",
"contribution_avg": 3.25,
"presentation": "2;3;3;3",
"presentation_avg": 2.75
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Spotlight",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.765191"
} | {
"id": "dToMLVhUEE",
"metareview": "All reviewers consider the proposed approach to text-to-image (T2I) model personalization using learnable low-rank parameters (PaRa) as innovative. They find the idea of controlling the rank of diffusion model parameters for personalization novel and well-explained. The experim... | {
"decision": "Accept (Spotlight)"
} |
KaYXsoCxV7 | 2410.15732v1 | ViMoE: An Empirical Study of Designing Vision Mixture-of-Experts | {
"content": "## Abstract\n\nAbstract Mixture-of-Experts (MoE) models embody the divide-and-conquer concept and are a promising approach for increasing model capacity, demonstrating excellent scalability across multiple domains. In this paper, we integrate the MoE structure into the classic Vision Transformer (ViT), ... | [
{
"id": "dQ5DzzfHi2",
"initial_rating": 3,
"confidence": 3,
"soundness": 2,
"contribution": 1,
"presentation": 2,
"summary": "The paper integrates the MoE structure into the Vision Transformer (ViT) to create ViMoE and studies its application in image classification. MoE models are promi... | {
"rating": "3;3;3;3",
"rating_avg": 3,
"confidence": "4;5;4;3",
"confidence_avg": 4,
"soundness": "2;2;2;2",
"soundness_avg": 2,
"contribution": "1;2;2;1",
"contribution_avg": 1.5,
"presentation": "2;3;2;2",
"presentation_avg": 2.25
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:01.766975"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
Kb1bIuGuax | 2410.11985v1 | The Fair Language Model Paradox | {
"content": "## Abstract\n\nAbstract Large Language Models (LLMs) are widely deployed in real-world applications, yet little is known about their training dynamics at the token level. Evaluation typically relies on aggregated training loss, measured at the batch level, which overlooks subtle per-token biases arising... | [
{
"id": "IC24TqehB0",
"initial_rating": 5,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "The paper titled The Fair Language Model Paradox presents an investigation into token-level biases in Large Language Models (LLMs) induced by weight decay, a comm... | {
"rating": "3;5;5;6",
"rating_avg": 4.75,
"confidence": "4;3;3;4",
"confidence_avg": 3.5,
"soundness": "2;3;3;3",
"soundness_avg": 2.75,
"contribution": "2;2;3;2",
"contribution_avg": 2.25,
"presentation": "2;3;3;3",
"presentation_avg": 2.75
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:01.770879"
} | {
"id": "cqeMfXa6qa",
"metareview": "The paper investigates the impact of weight decay, a common regularization technique, on token-level learning dynamics in LLM. The key finding is that as weight decay increases, the performance of low-frequency tokens is disproportionately degraded, while high-frequency tokens a... | {
"decision": "Reject"
} |
Kdcqzfypry | 2408.10700v1 | AnyGraph: Graph Foundation Model in the Wild | {
"content": "## Abstract\n\nAbstract. The growing ubiquity of relational data structured as graphs has underscored the need for graph learning models with exceptional generalization capabilities. However, current approaches often struggle to effectively extract generalizable insights, frequently requiring extensive ... | [
{
"id": "NUB2Xbtyp0",
"initial_rating": 5,
"confidence": 4,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "The paper proposes a foundation graph model to handle graphs with (numerical) node attributes.\n\nIt used 38 diverse graph datasets for training.\n\nIt proposes a... | {
"rating": "3;3;5;5;5",
"rating_avg": 4.2,
"confidence": "4;4;4;4;4",
"confidence_avg": 4,
"soundness": "2;2;2;3;3",
"soundness_avg": 2.4,
"contribution": "2;2;2;3;2",
"contribution_avg": 2.2,
"presentation": "3;1;1;3;3",
"presentation_avg": 2.2
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:01.773896"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
Kg0KnPfiW2 | 2410.01784v1 | Automating Large-scale In-silico Benchmarking for Genomic Foundation Models | {
"content": "## 1 Introduction\n\nThe central dogma of biology [] posits that genomes, including DNA and RNA, encode and transmit the genetic information essential for all living systems and underpin the translation of proteins. Despite decades of advancements in molecular biology, deciphering genomes remains a sign... | [
{
"id": "RPvRP6u8Z5",
"initial_rating": 3,
"confidence": 3,
"soundness": 2,
"contribution": 2,
"presentation": 1,
"summary": "This paper introduces a meta-benchmark for transfer learning that integrates tasks from four existing benchmark sets. They revise metrics used in these benchmarks... | {
"rating": "1;3;3;5",
"rating_avg": 3,
"confidence": "4;3;3;3",
"confidence_avg": 3.25,
"soundness": "2;2;2;3",
"soundness_avg": 2.25,
"contribution": "1;1;2;3",
"contribution_avg": 1.75,
"presentation": "2;2;1;4",
"presentation_avg": 2.25
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:01.775769"
} | {
"id": "oFPMW5xhfc",
"metareview": "This paper is about developing a better benchmark for transfer learning tasks. The construct a benchmark that is a composition of existing benchmarks sets and implement it in python. Reviewers noted the value of benchmarking models so that the community can assess performance an... | {
"decision": "Reject"
} |
KhvBlzwehb | 2410.04659v1 | ActiView: Evaluating Active Perception Ability for Multimodal Large Language Models | {
"content": "## Abstract\n\nAbstract Active perception, a crucial human capability, involves setting a goal based on the current understanding of the environment and performing actions to achieve that goal. Despite significant efforts in evaluating Multimodal Large Language Models (MLLMs), active perception has been... | [
{
"id": "3tdkE0UFQZ",
"initial_rating": 5,
"confidence": 5,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "This paper takes an \"Active Perception\" perspective for multimodal LLMs and proposes an evaluation benchmark where VQA is chosen as the proxy task. Given an im... | {
"rating": "3;3;5;6",
"rating_avg": 4.25,
"confidence": "4;4;5;4",
"confidence_avg": 4.25,
"soundness": "2;2;3;3",
"soundness_avg": 2.5,
"contribution": "2;2;2;3",
"contribution_avg": 2.25,
"presentation": "3;3;3;3",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:01.777405"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
KlN00vQEY2 | 2410.05898v3 | Manifolds, Random Matrices and Spectral Gaps: The geometric phases of generative diffusion | {
"content": "## Abstract\n\nAbstract In this paper, we investigate the latent geometry of generative diffusion models under the manifold hypothesis. To this purpose, we analyze the spectrum of eigenvalues (and singular values) of the Jacobian of the score function, whose discontinuities (gaps) reveal the presence an... | [
{
"id": "4raC233v8n",
"initial_rating": 6,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "The paper studies the generative process of diffusion models, meaning how samples of the target distribution are generated starting from random initial values, by... | {
"rating": "3;5;5;6;6",
"rating_avg": 5,
"confidence": "3;4;3;3;4",
"confidence_avg": 3.4,
"soundness": "3;3;2;3;3",
"soundness_avg": 2.8,
"contribution": "2;2;3;2;3",
"contribution_avg": 2.4,
"presentation": "1;2;2;2;3",
"presentation_avg": 2
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.779274"
} | {
"id": "CQS0YGHa3S",
"metareview": "Summary:\nThis paper presents a novel theoretical analysis of generative diffusion models through the lens of differential geometry and random matrix theory. By examining the spectral gaps in the Jacobian of the score function, the authors identify three distinct phases of gener... | {
"decision": "Accept (Poster)"
} |
KlV5CkNQkl | 2403.15576v2 | Data-centric Prediction Explanation via Kernelized Stein Discrepancy | {
"content": "## Abstract\n\nAbstract Existing example-based prediction explanation methods often bridge test and training data points through the model’s parameters or latent representations. While these methods offer clues to the causes of model predictions, they often exhibit innate shortcomings, such as incurring... | [
{
"id": "IRLBaI0Jeq",
"initial_rating": 5,
"confidence": 3,
"soundness": 3,
"contribution": 2,
"presentation": 2,
"summary": "The paper presents HD-Explain, a data-centric, example-based explanation method leveraging Kernelized Stein Discrepancy (KSD). The authors argue that existing exa... | {
"rating": "5;6;6;6",
"rating_avg": 5.75,
"confidence": "3;3;3;3",
"confidence_avg": 3,
"soundness": "3;2;3;3",
"soundness_avg": 2.75,
"contribution": "2;3;3;3",
"contribution_avg": 2.75,
"presentation": "2;1;4;3",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.781147"
} | {
"id": "TGIDvlJ1r3",
"metareview": "The paper presents a framework for data-centric example-based model explanation using Kernelized Stein discrepancy. The main claimed contribution is that the proposed method outperforms previous work in terms of explanation precision, computational efficiency and consistency. Ov... | {
"decision": "Accept (Poster)"
} |
KlxK4ncqWZ | 2410.11275v1 | Shallow diffusion networks provably learn hidden low-dimensional structure | {
"content": "## Abstract\n\nAbstract Diffusion-based generative models provide a powerful framework for learning to sample from a complex target distribution.\nThe remarkable empirical success of these models applied to high-dimensional signals, including images and video, stands in stark contrast to classical resul... | [
{
"id": "G74fOg67WF",
"initial_rating": 6,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 4,
"summary": "This paper investigates why diffusion models, specifically those utilizing shallow neural networks within Barron spaces, can effectively learn and sample from hig... | {
"rating": "5;6;6;8",
"rating_avg": 6.25,
"confidence": "3;1;3;2",
"confidence_avg": 2.25,
"soundness": "4;2;3;3",
"soundness_avg": 3,
"contribution": "3;2;3;3",
"contribution_avg": 2.75,
"presentation": "3;2;4;1",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.782373"
} | {
"id": "Mq7vkwhdIz",
"metareview": "The paper is a solid mathematical analysis of an interesting phenomenon and will be of interest to much of the community. Some complexities, such as very deep models and algorithmic aspects, have been simplified for the sake of theoretical results, but even without them, the con... | {
"decision": "Accept (Poster)"
} |
KmphHE92wU | 2410.09737v1 | Towards Stable, Globally Expressive Graph Representations with Laplacian Eigenvectors | {
"content": "## Abstract\n\nAbstract Graph neural networks (GNNs) have achieved remarkable success in a variety of machine learning tasks over graph data. Existing GNNs usually rely on message passing, i.e., computing node representations by gathering information from the neighborhood, to build their underlying comp... | [
{
"id": "YdQxskwUmZ",
"initial_rating": 5,
"confidence": 2,
"soundness": 2,
"contribution": 3,
"presentation": 3,
"summary": "The paper presents a robust orthogonal group-equivariant neural network designed to enhance graph representation using Laplacian vectors. Experiments conducted on... | {
"rating": "5;5;6;6",
"rating_avg": 5.5,
"confidence": "4;2;4;4",
"confidence_avg": 3.5,
"soundness": "3;2;2;2",
"soundness_avg": 2.25,
"contribution": "2;3;2;3",
"contribution_avg": 2.5,
"presentation": "2;3;3;3",
"presentation_avg": 2.75
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:01.783891"
} | {
"id": "hhe3BjNHfp",
"metareview": "This paper addresses the problem of how to incorporate Laplacian eigenvectors into message-passing neural networks to maximize expressivity while maintaining invariance/equivariance. The authors propose two main architectures to achieve this based on positional encodings from La... | {
"decision": "Reject"
} |
KrK6zXbjfO | 2405.18503v2 | SoundCTM: Unifying Score-based and Consistency Models for Full-band Text-to-Sound Generation | {
"content": "## Abstract\n\nAbstract Sound content is an indispensable element for multimedia works such as video games, music, and films.\nRecent high-quality diffusion-based sound generation models can serve as valuable tools for the creators.\nHowever, despite producing high-quality sounds, these models often suf... | [
{
"id": "kpofCVL5Ww",
"initial_rating": 6,
"confidence": 5,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "This paper presents a diffusion-based text-to-audio generation system within the Consistency Trajectory Model (CTM) framework, named SoundCTM. SoundCTM is develop... | {
"rating": "3;6;6;6",
"rating_avg": 5.25,
"confidence": "5;3;2;5",
"confidence_avg": 3.75,
"soundness": "2;3;3;3",
"soundness_avg": 2.75,
"contribution": "2;2;3;3",
"contribution_avg": 2.5,
"presentation": "2;3;2;3",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.784898"
} | {
"id": "QjzVLliC04",
"metareview": "This paper proposes Sound Consistency Trajectory Models (SoundCTM), a text-to-sound generation framework that unifies score-based diffusion and consistency modeling. The core objective is to enable both fast trial-and-error via high-quality one-step sampling and refined multi-st... | {
"decision": "Accept (Poster)"
} |
KrSaWQH1OA | 2407.20801v1 | AhmedML: High-Fidelity Computational Fluid Dynamics Dataset for Incompressible, Low-Speed Bluff Body Aerodynamics | {
"content": "## Abstract\n\nAbstract The development of Machine Learning (ML) methods for Computational Fluid Dynamics (CFD) is currently limited by the lack of openly available training data. This paper presents a new open-source dataset comprising of high fidelity, scale-resolving CFD simulations of 500 geometric ... | [
{
"id": "05iIzk2wY1",
"initial_rating": 5,
"confidence": 4,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "The ML-based methods depends on the \"true\" solutions simulated by traditional CFD solvers. The high-quality dataset is lacking in the field of ML for CFD.\n\nTo... | {
"rating": "3;3;3;5;6",
"rating_avg": 4,
"confidence": "3;5;4;4;5",
"confidence_avg": 4.2,
"soundness": "1;3;3;3;3",
"soundness_avg": 2.6,
"contribution": "2;2;1;2;3",
"contribution_avg": 2,
"presentation": "1;3;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": "2026-01-14T22:16:01.785658"
} | {
"id": "i6xzuLjC8a",
"metareview": "This domain-specific dataset paper introduces AhmedML, a high-fidelity, open-source dataset for CFD designed to advance ML applications in aerodynamic simulations. The dataset provides a collection of simulations using a hybrid RANS-LES approach in OpenFOAM for 500 geometric var... | {
"decision": "Reject"
} |
KsVlV2CRya | 2403.05571v4 | DiffuSolve: Diffusion-Based Solver for Non-Convex Trajectory Optimization | {
"content": "## Abstract\n\nAbstract Optimal trajectory design is computationally expensive for nonlinear and high-dimensional dynamical systems.\nThe challenge arises from the non-convex nature of the optimization problem with multiple local optima, which usually requires a global search.\nTraditional numerical sol... | [
{
"id": "zh87e7VCed",
"initial_rating": 3,
"confidence": 2,
"soundness": 2,
"contribution": 2,
"presentation": 3,
"summary": "This paper proposes to use diffusion models (DMs) to enhance the nonlinear program (NLP) solvers for trajectory optimization. The authors first collect local-opti... | {
"rating": "3;3;3;6",
"rating_avg": 3.75,
"confidence": "3;4;2;2",
"confidence_avg": 2.75,
"soundness": "2;4;2;3",
"soundness_avg": 2.75,
"contribution": "2;1;2;3",
"contribution_avg": 2,
"presentation": "3;3;3;3",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:01.786339"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
KstDMYkfj4 | 2202.00563v2 | On the Limitations of General Purpose Domain Generalisation Methods | {
"content": "## Abstract\n\nAbstract We investigate the fundamental performance limitations of learning algorithms in several Domain Generalisation (DG) settings. Motivated by the difficulty with which previously proposed methods have in reliably outperforming Empirical Risk Minimisation (ERM), we derive upper bound... | [
{
"id": "h6GTTvlagc",
"initial_rating": 3,
"confidence": 4,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "The paper considers upper bounds for ERM, as well as min-max lower bounds for domain generalization, in settings in which we are given n total environments and m ... | {
"rating": "3;3;3;5;5",
"rating_avg": 3.8,
"confidence": "1;5;4;4;3",
"confidence_avg": 3.4,
"soundness": "2;1;3;3;2",
"soundness_avg": 2.2,
"contribution": "1;2;2;2;2",
"contribution_avg": 1.8,
"presentation": "3;3;3;3;2",
"presentation_avg": 2.8
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:01.787063"
} | {
"id": "wyrEWavcRi",
"metareview": "This paper investigates domain generalization (DG) from a learning theory perspective, focusing on the performance of ERM. The authors provide upper and lower bounds for the excess risk in DG settings, considering different assumptions about domain similarity, and conclude that ... | {
"decision": "Reject"
} |
Kvdh12wGC0 | 2410.14735v1 | Agent Skill Acquisition for Large Language Models via CycleQD | {
"content": "## Abstract\n\nAbstract Training large language models to acquire specific skills remains a challenging endeavor.\nConventional training approaches often struggle with data distribution imbalances and inadequacies in objective functions that do not align well with task-specific performance.\nTo address ... | [
{
"id": "hqepJjUeGG",
"initial_rating": 6,
"confidence": 1,
"soundness": 3,
"contribution": 3,
"presentation": 2,
"summary": "This paper presents CycleQD, a novel framework for making llm acquire specific skills through Quality Diversity (QD) optimization. CycleQD addresses challenges in... | {
"rating": "5;6;6;6",
"rating_avg": 5.75,
"confidence": "3;4;2;1",
"confidence_avg": 2.5,
"soundness": "2;3;3;3",
"soundness_avg": 2.75,
"contribution": "3;3;3;3",
"contribution_avg": 3,
"presentation": "1;3;3;2",
"presentation_avg": 2.25
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.787898"
} | {
"id": "oqzWD169Jx",
"metareview": "This paper introduces CycleQD, a framework for skill acquisition in large language models (LLMs) using Quality Diversity (QD) optimization. By cyclically alternating quality measures for different tasks, CycleQD offers a scalable alternative to fine-tuning and traditional model ... | {
"decision": "Accept (Poster)"
} |
KwPUQOQIKt | 2406.06592v1 | Improve Mathematical Reasoning in Language Models with Automated Process Supervision | {
"content": "## Abstract\n\nAbstract Complex multi-step reasoning tasks, such as solving mathematical problems or generating code, remain a significant hurdle for even the most advanced large language models (LLMs). Verifying LLM outputs with an Outcome Reward Model (ORM) is a standard inference-time technique aimed... | [
{
"id": "c2UMx1puZf",
"initial_rating": 6,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "This paper proposes to train a PRM as verifier for LLM math reasoning. Since human annotation for step-wise reasoning is expensive, the authors propose to use mct... | {
"rating": "3;5;6",
"rating_avg": 4.666666666666667,
"confidence": "4;4;3",
"confidence_avg": 3.6666666666666665,
"soundness": "1;3;3",
"soundness_avg": 2.3333333333333335,
"contribution": "2;2;3",
"contribution_avg": 2.3333333333333335,
"presentation": "1;2;3",
"presentation_avg": 2
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:01.788768"
} | {
"id": "2NXfQErwpu",
"metareview": "This paper proposes a new MCTS-based algorithm for the efficient collection of high-quality data for learning Process Reward Model (PRM), and show the improvements gained by proposed methods on existing benchmarks for math problems. While reviewers raised several questions conce... | {
"decision": "Reject"
} |
KxQRHOre9D | 2410.09644v1 | Adapters for Altering LLM Vocabularies: What Languages Benefit the Most? | {
"content": "## Abstract\n\nAbstract Vocabulary adaptation, which integrates new vocabulary into pre-trained language models (LMs), enables expansion to new languages and mitigates token over-fragmentation.\nHowever, existing approaches are limited by their reliance on heuristic or external embeddings.\nWe propose V... | [
{
"id": "l39EQnn2z2",
"initial_rating": 6,
"confidence": 5,
"soundness": 3,
"contribution": 3,
"presentation": 4,
"summary": "The paper introduces VocADT, an adapter-based method for vocabulary adaptation in pre-trained language models. New embeddings for an expanded vocabulary are learn... | {
"rating": "3;5;6;6",
"rating_avg": 5,
"confidence": "4;3;4;5",
"confidence_avg": 4,
"soundness": "2;2;3;3",
"soundness_avg": 2.5,
"contribution": "2;2;3;3",
"contribution_avg": 2.5,
"presentation": "1;3;3;4",
"presentation_avg": 2.75
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.789471"
} | {
"id": "T07QfVeJ9D",
"metareview": "This paper presents VocADT, a method for multilingual vocabulary adaptation using adapters. Specifically, rather than altering the embedding layer of the multilingual model to create a new multilingual vocabulary, the work proposes using adapters to learn linear combinations of ... | {
"decision": "Accept (Poster)"
} |
L07zWidgdW | 2405.17663v2 | Finding Shared Decodable Concepts and their Negations in the Brain | {
"content": "## Abstract\n\nAbstract Prior work has offered evidence for functional localization in the brain; different anatomical regions preferentially activate for certain types of visual input. For example, the fusiform face area preferentially activates for visual stimuli that include a face. However, the spec... | [
{
"id": "n7604Ru5IE",
"initial_rating": 5,
"confidence": 4,
"soundness": 2,
"contribution": 2,
"presentation": 4,
"summary": "This study aims to identify brain areas that respond to specific visual concepts, aiming to expand our understanding of functional localization beyond known areas... | {
"rating": "3;5;5;8",
"rating_avg": 5.25,
"confidence": "4;4;4;3",
"confidence_avg": 3.75,
"soundness": "3;3;2;4",
"soundness_avg": 3,
"contribution": "2;3;2;3",
"contribution_avg": 2.5,
"presentation": "2;3;4;3",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.790093"
} | {
"id": "cMQgD4w3Bk",
"metareview": "This paper introduces a contrastive model that maps brain responses during naturalistic image viewing to CLIP embeddings, revealing Shared Decodable Concepts (SDCs)—clusters of visual-semantic features linked to common brain regions. Using a novel adaptation of DBSCAN, the study... | {
"decision": "Accept (Poster)"
} |
L0PciKdHsP | 2410.09687v1 | MoIN: Mixture of Introvert Experts to Upcycle an LLM | {
"content": "## Abstract\n\nAbstract The goal of this paper is to improve (upcycle) an existing large language model without the prohibitive requirements of continued pre-training of the full-model. The idea is to split the pre-training data into semantically relevant groups and train an expert on each subset. An ex... | [
{
"id": "YV4P2DOPs6",
"initial_rating": 6,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "The paper introduces a novel way to upcycle a pre-trained language model, allowing improvement without the high costs of full model pre-training. The paper utiliz... | {
"rating": "3;3;6;6",
"rating_avg": 4.5,
"confidence": "5;4;2;3",
"confidence_avg": 3.5,
"soundness": "2;1;2;3",
"soundness_avg": 2,
"contribution": "1;1;2;3",
"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": "2026-01-14T22:16:01.790741"
} | {
"id": "TRhOtElk0d",
"metareview": "The paper presents an alternative to the building of a single LLM, through designing a host of light LLMs dedicated to different topics; each ,light LLM is defined as a LoRA adapter, and queries are routed to the appropriate expert.\nThe strategy can be summarized as \"creating ... | {
"decision": "Reject"
} |
L0pXYjtfE3 | 2406.17296v1 | BlockLLM: Memory-Efficient Adaptation of LLMs by Selecting and Optimizing the Right Coordinate Blocks | {
"content": "## Abstract\n\nAbstract Training large language models (LLMs) for pretraining or adapting to new tasks and domains has become increasingly critical as their applications expand. However, as the model and the data sizes grow, the training process presents significant memory challenges, often requiring a ... | [
{
"id": "SUFI9d7sRj",
"initial_rating": 3,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 2,
"summary": "In this work, the authors introduce BlockLLM, a novel approach for memory-efficient LLM pre-training and adaptation. The method achieves efficiency by selectively... | {
"rating": "3;3;3;5;5",
"rating_avg": 3.8,
"confidence": "4;4;3;3;3",
"confidence_avg": 3.4,
"soundness": "2;1;3;3;3",
"soundness_avg": 2.4,
"contribution": "2;2;3;3;2",
"contribution_avg": 2.4,
"presentation": "2;3;2;3;2",
"presentation_avg": 2.4
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:01.791329"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
L14sqcrUC3 | 2406.19380v4 | Analyzing Pitfalls and Filling the Gaps in Tabular Deep Learning Benchmarks | {
"content": "## Abstract\n\nAbstract Advances in machine learning research drive progress in real-world applications.\nTo ensure this progress, it is important to understand the potential pitfalls on the way from a novel method’s success on academic benchmarks to its practical deployment. In this work, we analyze ex... | [
{
"id": "wOb6DKt0Gn",
"initial_rating": 8,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "The present paper studies the characteristics of academic tabular benchmarks for deep learning models, and makes the following contributions:\n\n- The authors ide... | {
"rating": "3;5;6;8",
"rating_avg": 5.5,
"confidence": "4;4;3;3",
"confidence_avg": 3.5,
"soundness": "2;2;2;4",
"soundness_avg": 2.5,
"contribution": "1;3;3;4",
"contribution_avg": 2.75,
"presentation": "3;3;3;3",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Spotlight",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.791996"
} | {
"id": "h83pT6NEka",
"metareview": "Tabular learning with temporal drift is an interesting problem with many use cases in practice and is hardly discussed or addressed in academic literature. The authors nicely summarise the related work in the paper which is generally well written and easy to understand. The clai... | {
"decision": "Accept (Spotlight)"
} |
L4nH3j7L94 | 2410.02712v1 | LLaVA-Critic: Learning to Evaluate Multimodal Models | {
"content": "## Abstract\n\nAbstract We introduce LLaVA-Critic, the first open-source large multimodal model (LMM) designed as a generalist evaluator to assess performance across a wide range of multimodal tasks. LLaVA-Critic is trained using a high-quality critic instruction-following dataset that incorporates dive... | [
{
"id": "8oxv5uepUq",
"initial_rating": 5,
"confidence": 3,
"soundness": 2,
"contribution": 2,
"presentation": 2,
"summary": "This paper proposes a high-quality dataset for evaluation purposes. Based on this dataset, it trains an MLLM mainly for evaluation use. The results show that LLAV... | {
"rating": "3;5;5;6",
"rating_avg": 4.75,
"confidence": "3;4;3;3",
"confidence_avg": 3.25,
"soundness": "3;3;2;2",
"soundness_avg": 2.5,
"contribution": "2;2;2;2",
"contribution_avg": 2,
"presentation": "2;3;2;3",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:01.792714"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
L5NUDBdHqR | 2406.04949v1 | Nacala-Roof-Material: Drone Imagery for Roof Detection, Classification, and Segmentation to Support Mosquito-borne Disease Risk Assessment | {
"content": "## Abstract\n\nAbstract As low-quality housing and in particular certain roof characteristics are associated with an increased risk of malaria, classification of roof types based on remote sensing imagery can support the assessment of malaria risk and thereby help prevent the disease. To support researc... | [
{
"id": "BRyynSoa2G",
"initial_rating": 5,
"confidence": 4,
"soundness": 4,
"contribution": 2,
"presentation": 3,
"summary": "This paper introduces the Nacala-Roof-Material dataset, a high-resolution drone imagery dataset collected from informal settlements in Nacala, Mozambique. The dat... | {
"rating": "1;3;5;5",
"rating_avg": 3.5,
"confidence": "4;5;3;4",
"confidence_avg": 4,
"soundness": "3;2;3;4",
"soundness_avg": 3,
"contribution": "1;2;2;2",
"contribution_avg": 1.75,
"presentation": "3;3;3;3",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:01.793991"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
L7gyAKWpiM | 2410.16542v2 | A Theoretical Study of Neural Network Expressive Power via Manifold Topology | {
"content": "## Abstract\n\nAbstract A prevalent assumption regarding real-world data is that it lies on or close to a low-dimensional manifold. When deploying a neural network on data manifolds, the required size, i.e., the number of neurons of the network, heavily depends on the intricacy of the underlying latent ... | [
{
"id": "dZZzpWjD6z",
"initial_rating": 5,
"confidence": 3,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "This paper explores the relationship between neural network size and the topology and geometry of data manifolds, providing a theoretical framework that integrate... | {
"rating": "5;5;6;6;6",
"rating_avg": 5.6,
"confidence": "3;3;2;3;3",
"confidence_avg": 2.8,
"soundness": "2;3;3;4;3",
"soundness_avg": 3,
"contribution": "2;2;3;2;2",
"contribution_avg": 2.2,
"presentation": "3;3;3;3;4",
"presentation_avg": 3.2
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:01.795079"
} | {
"id": "REPRr4H7xW",
"metareview": "The current paper explored how the network size scales up with respect to the topological complexity and curvature of the data manifold via a constructive proof. All reviewers acknowledged that the theoretical contribution is novel and interesting, and that the authors took a pr... | {
"decision": "Reject"
} |
L8e7tBf4pP | 2410.03782v1 | DaWin: Training-free Dynamic Weight Interpolation for Robust Adaptation | {
"content": "## Abstract\n\nAbstract Adapting a pre-trained foundation model on downstream tasks should ensure robustness against distribution shifts without the need to retrain the whole model. Although existing weight interpolation methods are simple yet effective, we argue their static nature limits downstream pe... | [
{
"id": "wdeUDdfU82",
"initial_rating": 6,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 4,
"summary": "A new method for robust fine-tuning, called \"dynamic weight interpolation\" (DaWin) is proposed. It is directly related to the well known model averaging method ... | {
"rating": "5;6;6;6",
"rating_avg": 5.75,
"confidence": "4;3;3;3",
"confidence_avg": 3.25,
"soundness": "3;3;3;3",
"soundness_avg": 3,
"contribution": "2;3;3;3",
"contribution_avg": 2.75,
"presentation": "3;3;3;4",
"presentation_avg": 3.25
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.796046"
} | {
"id": "gSqjWBjxkU",
"metareview": "This submission introduces DaWin, a dynamic weight interpolation method that combines pre-trained and fine-tuned models based on per-sample entropy without requiring additional training. It further utilizes Beta mixture models to efficiently model interpolation coefficients, ens... | {
"decision": "Accept (Poster)"
} |
L9eEfwwUwU | 2411.06770v2 | Sketched Adaptive Federated Deep Learning: A Sharp Convergence Analysis | {
"content": "## Abstract\n\nAbstract Combining gradient compression methods (e.g., CountSketch (Charikar et al.,, 2002 ; Rothchild et al.,, 2020 ) , quantization (Tang et al.,, 2021 ) ) and adaptive optimizers (e.g., Adam (Kingma and Ba,, 2014 ) , AMSGrad (Reddi et al.,, 2019 ) ) is a desirable goal in federated lea... | [
{
"id": "dWIrYk6Hdx",
"initial_rating": 3,
"confidence": 4,
"soundness": 2,
"contribution": 2,
"presentation": 3,
"summary": "This paper introduces the Sketched Adaptive Federated Learning (SAFL) framework, aiming to alleviate communication burdens in federated learning (FL) by leveragin... | {
"rating": "3;6;6",
"rating_avg": 5,
"confidence": "4;4;3",
"confidence_avg": 3.6666666666666665,
"soundness": "3;3;4",
"soundness_avg": 3.3333333333333335,
"contribution": "1;3;3",
"contribution_avg": 2.3333333333333335,
"presentation": "4;2;3",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:01.797303"
} | {
"id": "d2nyvaQ9U5",
"metareview": "This paper introduces the Sketched Adaptive Federated Learning (SAFL) framework, aiming to alleviate communication burdens in federated learning (FL) by leveraging gradient sketching alongside adaptive optimizers. The authors propose that by combining techniques like CountSketch... | {
"decision": "Reject"
} |
L9j8exYGUJ | 2406.13858v1 | Distributional reasoning in LLMs: Parallel Reasoning Processes in Multi-hop Reasoning | {
"content": "## Abstract\n\nAbstract Large language models (LLMs) have shown an impressive ability to perform tasks believed to require \"thought processes”. When the model does not document an explicit thought process, it becomes difficult to understand the processes occurring within its hidden layers and to determ... | [
{
"id": "QwORTXwFUX",
"initial_rating": 5,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "This paper introduces a method to analyze multi-hop reasoning in LLMs by modeling their prediction process as a linear transformation between semantic spaces. The... | {
"rating": "3;5;6;6",
"rating_avg": 5,
"confidence": "3;4;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;4",
"presentation_avg": 3.25
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:01.798197"
} | {
"id": "fcXqyh5HJt",
"metareview": "The paper investigates the internal multi-hop reasoning processes in large language models (LLMs). The authors propose that LLMs utilize a \"distributional reasoning mechanism\", whereby intermediate layers generate embeddings representing potential intermediate answers, which a... | {
"decision": "Reject"
} |
LAsMFAg4Zf | 2310.04055v4 | Kick Bad Guys Out! Conditionally Activated Anomaly Detection in Federated Learning with Zero-Knowledge Proof Verification | {
"content": "## Abstract\n\nAbstract Federated Learning (FL) systems are susceptible to adversarial attacks, where malicious clients submit poisoned models to disrupt the convergence or plant backdoors that cause the global model to misclassify some samples. Current defense methods are often impractical for real-wor... | [
{
"id": "xuKPClip95",
"initial_rating": 3,
"confidence": 4,
"soundness": 2,
"contribution": 2,
"presentation": 3,
"summary": "The paper presents a defense against poisoning attacks in federated learning settings capable of detecting and removing the effect of malicious participants. The ... | {
"rating": "3;3;3;6",
"rating_avg": 3.75,
"confidence": "4;4;4;4",
"confidence_avg": 4,
"soundness": "2;2;2;3",
"soundness_avg": 2.25,
"contribution": "2;2;2;3",
"contribution_avg": 2.25,
"presentation": "3;3;3;3",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:01.798996"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
LB5cKhgOTu | 2410.06040v1 | QERA: an Analytical Framework for Quantization Error Reconstruction | {
"content": "## Abstract\n\nAbstract The growing number of parameters and computational demands of large language models (LLMs) present significant challenges for their efficient deployment.\nRecently, there is an increasing interest in quantizing weights to extremely low precision while offsetting the resulting err... | [
{
"id": "jFYgY7zxuT",
"initial_rating": 5,
"confidence": 4,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "This work proposes a framework for analyzing the quantization error after it is compensated using low-rank, high-precision terms. In such a decomposition, a weigh... | {
"rating": "3;5;5;6;8",
"rating_avg": 5.4,
"confidence": "4;4;4;4;3",
"confidence_avg": 3.8,
"soundness": "2;2;3;3;4",
"soundness_avg": 2.8,
"contribution": "2;2;2;3;3",
"contribution_avg": 2.4,
"presentation": "2;3;3;3;3",
"presentation_avg": 2.8
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.799760"
} | {
"id": "j3Wnl54D6g",
"metareview": "Dear Authors,\n\nThank you for your valuable contribution to the ICLR and the ML community. Your submitted paper has undergone a rigorous review process, and I have carefully read and considered the feedback provided by the reviewers.\n\nThis paper proposes a framework for analy... | {
"decision": "Accept (Poster)"
} |
LC2KxRwC3n | 2409.14507v4 | A is for Absorption: Studying Feature Splitting and Absorption in Sparse Autoencoders | {
"content": "## Abstract\n\nAbstract Sparse Autoencoders (SAEs) have emerged as a promising approach to decompose the activations of Large Language Models (LLMs) into human-interpretable latents. In this paper, we pose two questions. First, to what extent do SAEs extract monosemantic and interpretable latents? Secon... | [
{
"id": "bljzMQT9we",
"initial_rating": 8,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "This paper explores the use of Sparse Autoencoders (SAEs) to decompose activations of Large Language Models (LLMs) into interpretable latent features. It addresse... | {
"rating": "3;6;8;8",
"rating_avg": 6.25,
"confidence": "3;4;3;3",
"confidence_avg": 3.25,
"soundness": "2;2;3;3",
"soundness_avg": 2.5,
"contribution": "2;3;4;3",
"contribution_avg": 3,
"presentation": "2;3;3;3",
"presentation_avg": 2.75
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:01.800462"
} | {
"id": "csFPJs2qPd",
"metareview": "Sparse Autoencoders (SAEs) have become a popular tool for interpreting neuron activation behaviors in large language models (LLMs). This paper questions the effectiveness of sparsity regularization in SAEs by identifying the phenomenon of feature absorption—a behavior where over... | {
"decision": "Reject"
} |
LCL8SMGxDY | 2402.06855v2 | For Better or For Worse? Learning Minimum Variance Features With Label Augmentation | {
"content": "## Abstract\n\nAbstract Data augmentation has been pivotal in successfully training deep learning models on classification tasks over the past decade. An important subclass of data augmentation techniques - which includes both label smoothing and Mixup - involves modifying not only the input data but al... | [
{
"id": "LPZKVJh2jm",
"initial_rating": 5,
"confidence": 3,
"soundness": 2,
"contribution": 3,
"presentation": 3,
"summary": "This manuscript studies the relationship between label augmentation techniques (label smoothing and Mixup) and feature variance. The authors demonstrate that labe... | {
"rating": "3;5;5;5;6",
"rating_avg": 4.8,
"confidence": "2;3;3;3;5",
"confidence_avg": 3.2,
"soundness": "3;3;3;2;3",
"soundness_avg": 2.8,
"contribution": "2;3;2;3;3",
"contribution_avg": 2.6,
"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": "2026-01-14T22:16:01.801208"
} | {
"id": "3msiXRaIle",
"metareview": "The paper studies the effect of label manipulation in classification tasks. It achieves formal results on the effect of two common manipulations, label smoothing and mixup, on the variance of features that are learnt for linear binary classifiers. Then, it studies non-linear mod... | {
"decision": "Accept (Poster)"
} |
LCk3umTAXx | 2410.04038v2 | Gamified crowd-sourcing of high-quality data for visual fine-tuning | {
"content": "## Abstract\n\nAbstract This paper introduces gamified adversarial prompting (GAP),\na framework that crowd-sources high-quality data for visual instruction tuning of large multimodal models.\nGAP transforms the data collection process into an engaging game,\nincentivizing players to provide fine-graine... | [
{
"id": "p0Io86JFo2",
"initial_rating": 6,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "This paper introduces the Gamified Adversarial Prompting (GAP) framework, aimed at enhancing the performance of multimodal AI models in visual question answering ... | {
"rating": "3;5;6;6",
"rating_avg": 5,
"confidence": "4;3;3;4",
"confidence_avg": 3.5,
"soundness": "2;3;3;3",
"soundness_avg": 2.75,
"contribution": "2;2;3;3",
"contribution_avg": 2.5,
"presentation": "2;3;2;3",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:01.801971"
} | {
"id": "MFTftPgthM",
"metareview": "The paper introduces the Gamified Adversarial Prompting (GAP) framework, aimed at enhancing the performance of multimodal AI models in visual question answering (VQA) tasks. Preliminary results indicate that this approach shows promise.\n\nOverall, the paper is well-written, and... | {
"decision": "Reject"
} |
LDAj4UJ4aL | 2410.03478v1 | VEDIT: Latent Prediction Architecture For Procedural Video Representation Learning | {
"content": "## Abstract\n\nAbstract Procedural video representation learning is an active research area where the objective is to learn an agent which can anticipate and forecast the future given the present video input, typically in conjunction with textual annotations.\nPrior works often rely on large-scale pretr... | [
{
"id": "Ew7Y0FbIhs",
"initial_rating": 5,
"confidence": 3,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "This paper introduces a novel pipeline,VEDIT, for procedural classifciation task using diffusion transformers without large-scale pre-training or additional langu... | {
"rating": "5;5;6;6;6",
"rating_avg": 5.6,
"confidence": "4;3;3;4;4",
"confidence_avg": 3.6,
"soundness": "2;3;3;3;2",
"soundness_avg": 2.6,
"contribution": "2;2;3;3;2",
"contribution_avg": 2.4,
"presentation": "2;3;3;3;2",
"presentation_avg": 2.6
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.802670"
} | {
"id": "RkpJPNLggq",
"metareview": "This paper receives final scores of 6,6,6,6,6, from five reviewers. Reviewers have raised several issues on the architectural novelty, scaling, and lack of detailed experimental analysis. The authors have addressed most of the concerns in the rebuttal. Thus, there is a consensus... | {
"decision": "Accept (Poster)"
} |
LGafQ1g2D2 | 2410.05440v2 | Can LLMs Understand Time Series Anomalies? | {
"content": "## Abstract\n\nAbstract Large Language Models (LLMs) have gained popularity in time series forecasting, but their potential for anomaly detection remains largely unexplored. Our study investigates whether LLMs can understand and detect anomalies in time series data, focusing on zero-shot and few-shot sc... | [
{
"id": "Yr9Daa7eWJ",
"initial_rating": 6,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "This paper studies how Large Language Models (LLMs) reason about anomalous time series. The authors formulate 7 hypothesis, based on recent work at the intersecti... | {
"rating": "3;3;5;6;6",
"rating_avg": 4.6,
"confidence": "5;4;4;4;4",
"confidence_avg": 4.2,
"soundness": "1;3;2;3;3",
"soundness_avg": 2.4,
"contribution": "1;3;2;3;3",
"contribution_avg": 2.4,
"presentation": "2;3;3;3;3",
"presentation_avg": 2.8
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.803424"
} | {
"id": "p5Gkk8wN6I",
"metareview": "This paper has been evaluated by 5 knowledgeable reviewers. Their opinions varied: 3 marginal acceptances, 1 marginal rejection and one straight rejection. The paper focuses on a timely topic and offers a broad set of hypotheses which are one by one more-or-less completely evalu... | {
"decision": "Accept (Poster)"
} |
LIBLIlk5M9 | 2409.07025v1 | CPSample: Classifier Protected Sampling for Guarding Training Data During Diffusion | {
"content": "## Abstract\n\nAbstract Diffusion models have a tendency to exactly replicate their training data, especially when trained on small datasets. Most prior work has sought to mitigate this problem by imposing differential privacy constraints or masking parts of the training data, resulting in a notable sub... | [
{
"id": "HjIH1Costw",
"initial_rating": 5,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "The paper proposes Classifier-Protected Sampling (CPSample), a novel technique to prevent diffusion models from generating training data duplicates, apart from p... | {
"rating": "3;5;5;5",
"rating_avg": 4.5,
"confidence": "4;4;4;4",
"confidence_avg": 4,
"soundness": "1;3;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": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.804154"
} | {
"id": "GHpjelJKSj",
"metareview": "This paper studies the problem of alleviating privacy leakage in diffusion models, which occurs when these reproduce samples very close to those in training data and membership inference attacks. This is an important problem, given that state-of-the-art models are often trained ... | {
"decision": "Accept (Poster)"
} |
LJGY2GVcit | 2410.01483v1 | Foldable SuperNets: Scalable Merging of Transformers with Different Initializations and Tasks | {
"content": "## Abstract\n\nAbstract Many recent methods aim to merge neural networks (NNs) with identical architectures trained on different tasks to obtain a single multi-task model. Most existing works tackle the simpler setup of merging NNs initialized from a common pre-trained network, where simple heuristics l... | [
{
"id": "DGvLdVfxxc",
"initial_rating": 6,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "This paper addresses the challenging problem of merging large transformers trained on different tasks from distinct initializations, where prior works typically r... | {
"rating": "3;5;6;6",
"rating_avg": 5,
"confidence": "3;4;3;3",
"confidence_avg": 3.25,
"soundness": "2;2;3;3",
"soundness_avg": 2.5,
"contribution": "1;2;3;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": "2026-01-14T22:16:01.805026"
} | {
"id": "8x4PcqurMu",
"metareview": "Summary: The paper introduces FS-Merge, a method for merging large transformers trained on different tasks with distinct initializations. It extends the concept of folding from previous work and uses feature reconstruction loss with unlabeled data for optimization. The method sh... | {
"decision": "Reject"
} |
LLWj8on4Rv | 2312.08558v1 | Leveraging Driver Field-of-View for Multimodal Ego-Trajectory Prediction | {
"content": "## Abstract\n\nAbstract Understanding the decision-making process of drivers is one of the keys to ensuring road safety.\nWhile the driver intent and the resulting ego-motion trajectory is valuable in developing driver-assistance systems, existing methods mostly focus on the motions of other vehicles.\n... | [
{
"id": "nAGQWWhBLx",
"initial_rating": 6,
"confidence": 5,
"soundness": 3,
"contribution": 3,
"presentation": 4,
"summary": "The paper introduces a multimodal ego-trajectory prediction model that utilizes driver field-of-view (FOV) data—including first-person video and gaze fixations—al... | {
"rating": "6;6;6",
"rating_avg": 6,
"confidence": "4;5;5",
"confidence_avg": 4.666666666666667,
"soundness": "3;4;3",
"soundness_avg": 3.3333333333333335,
"contribution": "3;3;3",
"contribution_avg": 3,
"presentation": "3;3;4",
"presentation_avg": 3.3333333333333335
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.807049"
} | {
"id": "D44Yvo3act",
"metareview": "The present work proposes a multimodal ego-motion prediction framework that considers a driver's field of view. It essentially has three contributions: the prediction network itself, a new metric for measuring trajectory complexity, and a dataset involving multicamera footage of... | {
"decision": "Accept (Poster)"
} |
LLtUtzSOL5 | 2410.08133v1 | Assessing Episodic Memory in LLMs with Sequence Order Recall Tasks | {
"content": "## Abstract\n\nAbstract Current LLM benchmarks focus on evaluating models’ memory of facts and semantic relations, primarily assessing semantic aspects of long-term memory. However, in humans, long-term memory also includes episodic memory, which links memories to their contexts, such as the time and pl... | [
{
"id": "7IQyc7IXow",
"initial_rating": 3,
"confidence": 4,
"soundness": 2,
"contribution": 2,
"presentation": 4,
"summary": "The authors propose a new dataset, called SORT, for assessing episodic memory in LLMs, centered around the task of recalling the correct order of two short book s... | {
"rating": "3;3;5;8",
"rating_avg": 4.75,
"confidence": "4;4;4;4",
"confidence_avg": 4,
"soundness": "2;2;2;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": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:01.807746"
} | {
"id": "jRvmtwA1fo",
"metareview": "This submission assesses the abilities of a set of language models (LMs) to perform \"sequence-order recall\", that is, recalling the order of elements of a previously seen sequence. The authors describe the capacity that performs this task in LMs \"episodic memory\", and give a... | {
"decision": "Reject"
} |
LNL7zKvm7e | 2410.03226v2 | Frame-Voyager: Learning to Query Frames for Video Large Language Models | {
"content": "## Abstract\n\nAbstract Video Large Language Models (Video-LLMs) have made remarkable progress in video understanding tasks.\nHowever, they are constrained by the maximum length of input tokens, making it impractical to input entire videos.\nExisting frame selection approaches, such as uniform frame sam... | [
{
"id": "eBwdjhho5d",
"initial_rating": 5,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 2,
"summary": "This paper proposed Frame-Voyager that learns to query informative frame combinations, based on the given textual queries in the task.\nAuthors introduced a new d... | {
"rating": "5;5;5;6",
"rating_avg": 5.25,
"confidence": "4;5;4;5",
"confidence_avg": 4.5,
"soundness": "1;2;3;3",
"soundness_avg": 2.25,
"contribution": "2;2;3;3",
"contribution_avg": 2.5,
"presentation": "3;3;2;3",
"presentation_avg": 2.75
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.808575"
} | {
"id": "n0WS54Qkhk",
"metareview": "Frame selection is a popular paradigm to improve the computational cost of video-language models. This paper improves upon prior works like Sevila by not choosing frames independently, but by considering temporal relations among frames (the authors address the combinatorial sear... | {
"decision": "Accept (Poster)"
} |
LNYIUouhdt | 2403.07378v4 | SVD-LLM: Truncation-aware Singular Value Decomposition for Large Language Model Compression | {
"content": "## Abstract\n\nAbstract The advancements in Large Language Models (LLMs) have been hindered by their substantial sizes, which necessitate LLM compression methods for practical deployment. Singular Value Decomposition (SVD) offers a promising solution for LLM compression.\nHowever, state-of-the-art SVD-b... | [
{
"id": "zwBXA26YpY",
"initial_rating": 5,
"confidence": 4,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "The authors investigate the compression of large language models through low rank approximation. They point out that the existing works in this direction have the... | {
"rating": "5;5;6;8",
"rating_avg": 6,
"confidence": "4;4;4;4",
"confidence_avg": 4,
"soundness": "3;3;3;4",
"soundness_avg": 3.25,
"contribution": "3;2;3;4",
"contribution_avg": 3,
"presentation": "2;3;4;4",
"presentation_avg": 3.25
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.809249"
} | {
"id": "yAjjbxAhsT",
"metareview": "This paper proposes a low-rank compression algorithm for compressing a weight matrix in a neural network. The algorithm is truncation aware and is better than naive SVD. Experimental results show promising performance gain over existing low-rank compression algorithms on LLMs. \... | {
"decision": "Accept (Poster)"
} |
LNkMWCEssX | 2406.09564v2 | Towards Domain Adaptive Neural Contextual Bandits | {
"content": "## Abstract\n\nAbstract Contextual bandit algorithms are essential for solving real-world decision making problems. In practice, collecting a contextual bandit’s feedback from different domains may involve different costs. For example, measuring drug reaction from mice (as a source domain) and humans (a... | [
{
"id": "lTZ3kkk5jZ",
"initial_rating": 8,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "The manuscript introduces a domain adaption algorithm for contextual bandit. It further proves that the proposed method achieves a sub-linear regret bound. Empiri... | {
"rating": "6;6",
"rating_avg": 6,
"confidence": "3;3",
"confidence_avg": 3,
"soundness": "3;3",
"soundness_avg": 3,
"contribution": "3;3",
"contribution_avg": 3,
"presentation": "2;3",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.809999"
} | {
"id": "8gI8a0AWQN",
"metareview": "The reviewers generally appreciate the novel algorithm for domain adaptation in bandits. I only have one more thing to add. Domain adaptation ultimately is a form of distributional shift. Another approach to deal with is through distributional robustness (see Distributionally Ro... | {
"decision": "Accept (Poster)"
} |
LOiYxBcGA9 | 2402.09113v2 | How does Your RL Agent Explore? An Optimal Transport Analysis of Occupancy Measure Trajectories | {
"content": "## Abstract\n\nAbstract The rising successes of RL are propelled by combining smart algorithmic strategies and deep architectures to optimize the distribution of returns and visitations over the state-action space.\nA quantitative framework to compare the learning processes of these eclectic RL algorith... | [
{
"id": "ZCFKIbWBV0",
"initial_rating": 6,
"confidence": 3,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "The paper introduces a novel method to assess the difficulty of learning in sequential decision-making problems. This is achieved by conceptualizing the learning ... | {
"rating": "3;5;6",
"rating_avg": 4.666666666666667,
"confidence": "4;2;3",
"confidence_avg": 3,
"soundness": "2;2;3",
"soundness_avg": 2.3333333333333335,
"contribution": "2;2;2",
"contribution_avg": 2,
"presentation": "3;3;3",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:01.811143"
} | {
"id": "bFTMAuxoz3",
"metareview": "This paper introduces the Effort of Sequential Learning (ESL) and Optimal Movement Ratio (OMR) metrics to analyze the learning dynamics of reinforcement learning (RL) algorithms by examining the trajectories of state-action occupancy measures. The paper provides theoretical resu... | {
"decision": "Reject"
} |
LPG8pPSfQD | 2410.14803v3 | DistRL: An Asynchronous Distributed Reinforcement Learning Framework for On-Device Control Agent | {
"content": "## Abstract\n\nAbstract On-device control agents, especially on mobile devices, are responsible for operating mobile devices to fulfill users’ requests, enabling seamless and intuitive interactions. Integrating Multimodal Large Language Models (MLLMs) into these agents enhances their ability to understa... | [
{
"id": "EYPz789plQ",
"initial_rating": 6,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "The paper presents DistRL, an asynchronous distributed reinforcement learning framework designed explicitly for on-device control agents in mobile environments. D... | {
"rating": "5;6;6;8",
"rating_avg": 6.25,
"confidence": "3;4;4;4",
"confidence_avg": 3.75,
"soundness": "3;2;3;4",
"soundness_avg": 3,
"contribution": "3;3;3;3",
"contribution_avg": 3,
"presentation": "2;1;3;4",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.812215"
} | {
"id": "ln2hxBSaqb",
"metareview": "**summary**\n\nThe paper introduces DistRL, an asynchronous distributed RL framework designed for training on-device control agents in mobile environments. Combining centralized training with decentralized data collection, DistRL addresses challenges like delayed updates and non... | {
"decision": "Accept (Poster)"
} |
LRSspInlN5 | 2405.20771v1 | Towards Black-Box Membership Inference Attack for Diffusion Models | {
"content": "## Abstract\n\nAbstract Identifying whether an artwork was used to train a diffusion model is an important research topic, given the rising popularity of AI-generated art and the associated copyright concerns.\nThe work approaches this problem from the membership inference attack (MIA) perspective. We f... | [
{
"id": "4HjeDX1A8l",
"initial_rating": 6,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 2,
"summary": "To provide protection for the artworks and detect misuse of data, this paper proposes a black-box membership inference attack for diffusion models used in image g... | {
"rating": "5;5;6;6",
"rating_avg": 5.5,
"confidence": "2;4;2;4",
"confidence_avg": 3,
"soundness": "2;3;3;3",
"soundness_avg": 2.75,
"contribution": "3;2;3;3",
"contribution_avg": 2.75,
"presentation": "2;3;3;2",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:01.813273"
} | {
"id": "CFdDHeb59i",
"metareview": "The paper introduces a membership inference attack method that uses only the image-to-image variation API. In terms of overall score, the reviewers were mildly positive (2x) or mildly negative (2x). In terms of strength, the reviewers highlighted that the method was intuitive an... | {
"decision": "Reject"
} |
LS1VuhkReU | 2408.06502v1 | Prompt Recovery for Image Generation Models: A Comparative Study of Discrete Optimizers | {
"content": "## Abstract\n\nAbstract Recovering natural language prompts for image generation models, solely based on the generated images is a difficult discrete optimization problem.\nIn this work, we present the first head-to-head comparison of recent discrete optimization techniques for the problem of prompt inv... | [
{
"id": "OiD9ixnSFl",
"initial_rating": 3,
"confidence": 3,
"soundness": 3,
"contribution": 2,
"presentation": 4,
"summary": "This paper presents an analysis of existing prompt optimization methodologies, including PEZ, Greedy Coordinate Gradients, AutoDAN, Random Search, PRISM, and Capt... | {
"rating": "3;3;3;3",
"rating_avg": 3,
"confidence": "4;4;5;3",
"confidence_avg": 4,
"soundness": "1;2;2;3",
"soundness_avg": 2,
"contribution": "1;1;2;2",
"contribution_avg": 1.5,
"presentation": "3;2;3;4",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:01.813938"
} | {
"id": "uGIzal3UHC",
"metareview": "The paper presents an empirical analysis of prompt recovery methods in text-to-image models, showcasing both qualitative and quantitative results.\n\nThe paper was reviewed by four experts in the domain who acknowledged that the paper was well written and easy to follow (BePT, u... | {
"decision": "Reject"
} |
LTDtjrv02Y | 2410.22936v1 | Bringing NeRFs to the Latent Space: Inverse Graphics Autoencoder | {
"content": "## Abstract\n\nAbstract While pre-trained image autoencoders are increasingly utilized in computer vision, the application of inverse graphics in 2D latent spaces has been under-explored.\nYet, besides reducing the training and rendering complexity, applying inverse graphics in the latent space enables ... | [
{
"id": "nTOfIrBRTR",
"initial_rating": 6,
"confidence": 4,
"soundness": 3,
"contribution": 2,
"presentation": 4,
"summary": "This paper introduces the concept of 3D-awareness into the latent space of autoencoders, through learning a latent NeRF and an Inverse Graphics AutoEncoder (IG-AE... | {
"rating": "3;5;6",
"rating_avg": 4.666666666666667,
"confidence": "4;5;4",
"confidence_avg": 4.333333333333333,
"soundness": "2;2;3",
"soundness_avg": 2.3333333333333335,
"contribution": "2;2;2",
"contribution_avg": 2,
"presentation": "2;2;4",
"presentation_avg": 2.6666666666666665
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.814558"
} | {
"id": "kfwA0FNfjw",
"metareview": "The authors present a method to align a 2D image autoencoder with a 3D latent space. The authors demonstrate that the resulting latent space leads to better NVS quality with less 2D artifacts.\n\nThe reviewers converged to positively-leaning scores of 6/6/6 over the course of di... | {
"decision": "Accept (Poster)"
} |
LVmafig6Tk | 2410.12334v1 | Generalized Smooth Stochastic Variational Inequalities: Almost Sure Convergence and Convergence Rates | {
"content": "## Abstract\n\nAbstract This paper focuses on solving a stochastic variational inequality (SVI) problem under relaxed smoothness assumption for a class of structured non-monotone operators. The SVI problem has attracted significant interest in the machine learning community due to its immediate applicat... | [
{
"id": "CDDKfEtz8n",
"initial_rating": 3,
"confidence": 3,
"soundness": 2,
"contribution": 2,
"presentation": 2,
"summary": "This paper focuses on stochastic variational inequality (SVI) problems involving non-monotone operators. Applying clipping to projected gradient descent and extra... | {
"rating": "3;5;5;6",
"rating_avg": 4.75,
"confidence": "3;4;3;3",
"confidence_avg": 3.25,
"soundness": "2;3;3;3",
"soundness_avg": 2.75,
"contribution": "2;3;2;2",
"contribution_avg": 2.25,
"presentation": "2;3;3;2",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:01.815497"
} | {
"id": "kcDMkNkaOK",
"metareview": "The paper studies stochastic variational inequalities under generalized smoothness assumptions and for operators that are \"quasi-sharp.\" Although the problem class is presented as corresponding to \"structured non-monotone\" operators, the quasi-sharpness is quite a strong ass... | {
"decision": "Reject"
} |
LWeVVPuIx0 | 2311.00676v1 | Last-Iterate Convergence Properties of Regret-Matching Algorithms in Games | {
"content": "## Abstract\n\nAbstract Algorithms based on regret matching, specifically regret matching + (RM + ), and its variants are the most popular approaches for solving large-scale two-player zero-sum games in practice.\nUnlike algorithms such as optimistic gradient descent ascent, which have strong last-itera... | [
{
"id": "UdIYvdWklf",
"initial_rating": 6,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "The authors study last-iterate convergence for Regret Matching + (RM+) algorithm and its recent variants PRM+ (Predictive RM+), ExRM+ (Extragradient RM+) and Smo... | {
"rating": "6;6;6;8",
"rating_avg": 6.5,
"confidence": "4;3;3;4",
"confidence_avg": 3.5,
"soundness": "3;3;3;3",
"soundness_avg": 3,
"contribution": "3;2;3;3",
"contribution_avg": 2.75,
"presentation": "3;3;3;4",
"presentation_avg": 3.25
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.816672"
} | {
"id": "WUPNBgjVet",
"metareview": "A great paper in last-iterate convergence of learning in games!",
"additional_comments": "NA"
} | {
"decision": "Accept (Poster)"
} |
LXVZQpEb2y | 2410.02136v1 | Disentangled Representation Learning for Parametric Partial Differential Equations | {
"content": "## Abstract\n\nAbstract Neural operators (NOs) have demonstrated remarkable success in learning mappings between function spaces, serving as efficient approximators for the forward solutions of complex physical systems governed by partial differential equations (PDEs). However, while effective as black-... | [
{
"id": "WeQcL45OqU",
"initial_rating": 5,
"confidence": 3,
"soundness": 2,
"contribution": 2,
"presentation": 2,
"summary": "This paper presents a model based on a neural operator architecture, integrating an encoder-decoder structure to derive interpretable representations and address ... | {
"rating": "3;5;5;5",
"rating_avg": 4.5,
"confidence": "3;3;3;3",
"confidence_avg": 3,
"soundness": "2;2;2;2",
"soundness_avg": 2,
"contribution": "2;2;2;2",
"contribution_avg": 2,
"presentation": "1;3;2;2",
"presentation_avg": 2
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:01.817730"
} | {
"id": "f3zcki4SIJ",
"metareview": "This paper proposes a new variational hyper-operator that aims to disentangle physical factors of variation from the neural operator parameters used for solving parametric PDEs. The approach can handle both forward and inverse problems simultaneously. The experimental results de... | {
"decision": "Reject"
} |
LXlTdn9hY9 | 2409.09085v1 | HESSO: Towards Automatic Efficient and User Friendly Any Neural Network Training and Pruning | {
"content": "## Abstract\n\nAbstract Structured pruning is one of the most popular approaches to effectively compress the heavy deep neural networks (DNNs) into compact sub-networks while retaining the original network performance. The existing methods suffer from multi-stage procedures along with significant engine... | [
{
"id": "hBR5Z5qYsm",
"initial_rating": 6,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "This paper proposes Hybrid Efficient Structured Sparse Optimizer (HESSO), a structured pruning algorithm. In addition, for reliably identifying the indispensable ... | {
"rating": "3;3;6;8",
"rating_avg": 5,
"confidence": "5;4;3;3",
"confidence_avg": 3.75,
"soundness": "2;1;3;3",
"soundness_avg": 2.25,
"contribution": "1;2;3;3",
"contribution_avg": 2.25,
"presentation": "3;2;3;3",
"presentation_avg": 2.75
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:01.818447"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
LYHEY783Np | 2410.12327v1 | Neuron based Personality Trait Induction in Large Language Models | {
"content": "## Abstract\n\nAbstract Large language models (LLMs) have become increasingly proficient at simulating various personality traits, an important capability for supporting related applications ( e.g., role-playing). To further improve this capacity, in this paper, we present a neuron-based approach for pe... | [
{
"id": "Jw08QWC5m8",
"initial_rating": 5,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 2,
"summary": "The paper presents a neuron manipulation approach to inducing Big Five personality traits in LLMs without retraining or altering model parameters. The paper intro... | {
"rating": "5;6;6",
"rating_avg": 5.666666666666667,
"confidence": "4;3;3",
"confidence_avg": 3.3333333333333335,
"soundness": "2;3;3",
"soundness_avg": 2.6666666666666665,
"contribution": "2;4;3",
"contribution_avg": 3,
"presentation": "2;4;3",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.819163"
} | {
"id": "bZKYbjIIgB",
"metareview": "This paper proposes a neuron manipulation method for inducing Big Five personality traits in large language models (LLMs) without the need for retraining or modifying model parameters. It also introduces PersonalityBench, a large-scale generative dataset grounded in real-world s... | {
"decision": "Accept (Poster)"
} |
LYawG8YkPa | 2410.07093v1 | LaMP: Language-Motion Pretraining for Motion Generation, Retrieval, and Captioning | {
"content": "## Abstract\n\nAbstract Language plays a vital role in the realm of human motion.\nExisting methods have largely depended on CLIP text embeddings for motion generation, yet they fall short in effectively aligning language and motion due to CLIP’s pretraining on static image-text pairs.\nThis work introd... | [
{
"id": "jIyAJUYp7k",
"initial_rating": 6,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "In this work, the authors point out that existing text-motion related methods largely depend on the CLIP text embeddings, yet will fall short in effectively align... | {
"rating": "5;6;6;6",
"rating_avg": 5.75,
"confidence": "3;4;4;4",
"confidence_avg": 3.75,
"soundness": "3;3;3;3",
"soundness_avg": 3,
"contribution": "2;3;3;3",
"contribution_avg": 2.75,
"presentation": "3;3;3;3",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.819921"
} | {
"id": "N4QAilbf8Q",
"metareview": "Previous methods for text-motion have relied primarily on CLIP text embedding for motion generation. However, this paper points out that CLIP is insufficient for effectively adjusting language and motion because it is pre-trained on pairs of static images and text. This study in... | {
"decision": "Accept (Poster)"
} |
LZfjxvqw0N | 2410.01699v1 | Accelerating Auto-regressive Text-to-Image Generation with Training-free Speculative Jacobi Decoding | {
"content": "## Abstract\n\nAbstract The current large auto-regressive models can generate high-quality, high-resolution images, but these models require hundreds or even thousands of steps of next-token prediction during inference, resulting in substantial time consumption. In existing studies, Jacobi decoding, an ... | [
{
"id": "KDXNOgjpC6",
"initial_rating": 6,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 2,
"summary": "This paper proposes Speculative Jacobi Decoding (SJD) for accelerating the inference of auto-regressive (AR) image generation. SJD introduces a probabilistic acce... | {
"rating": "6;6;6",
"rating_avg": 6,
"confidence": "2;3;4",
"confidence_avg": 3,
"soundness": "2;3;3",
"soundness_avg": 2.6666666666666665,
"contribution": "2;2;3",
"contribution_avg": 2.3333333333333335,
"presentation": "2;3;2",
"presentation_avg": 2.3333333333333335
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.820912"
} | {
"id": "zMMITrCi4g",
"metareview": "This paper proposes Speculative Jacobi Decoding, a training-free probabilistic parallel decoding algorithm for accelerating auto-regressive text-to-image generation. The core novelty lies in introducing a probabilistic convergence criterion, which allows for faster decoding with... | {
"decision": "Accept (Poster)"
} |
Lb91pXwZMR | 2410.10516v1 | UniGEM: A Unified Approach to Generation and Property Prediction for Molecules | {
"content": "## Abstract\n\nAbstract Molecular generation and molecular property prediction are both crucial for drug discovery, but they are often developed independently. Inspired by recent studies, which demonstrate that diffusion model, a prominent generative approach, can learn meaningful data representations t... | [
{
"id": "0zrarm0pON",
"initial_rating": 5,
"confidence": 4,
"soundness": 2,
"contribution": 2,
"presentation": 3,
"summary": "This paper proposes a unified model for molecular generation and property predictions. Since diffusion models require diffusing the 3D molecular conformers to a c... | {
"rating": "5;5;6",
"rating_avg": 5.333333333333333,
"confidence": "4;4;4",
"confidence_avg": 4,
"soundness": "3;2;3",
"soundness_avg": 2.6666666666666665,
"contribution": "2;2;3",
"contribution_avg": 2.3333333333333335,
"presentation": "3;3;3",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.821673"
} | {
"id": "ZzD3Tltgmc",
"metareview": "The paper proposes a novel pretraining approach for molecule property prediction and generation that aims to better align two main types of pretraining, namely prediction-based and generation-base.\n\nReviewers found the proposal of integration of generation and prediction based... | {
"decision": "Accept (Poster)"
} |
LbEWwJOufy | 2410.04221v1 | TANGO: Co-Speech Gesture Video Reenactment with Hierarchical Audio Motion Embedding and Diffusion Interpolation | {
"content": "## Abstract\n\nAbstract We present TANGO, a framework for generating co-speech body-gesture videos. Given a few-minute, single-speaker reference video and target speech audio, TANGO produces high-fidelity videos with synchronized body gestures. TANGO builds on Gesture Video Reenactment (GVR), which spli... | [
{
"id": "FXR5DbPidQ",
"initial_rating": 8,
"confidence": 5,
"soundness": 4,
"contribution": 4,
"presentation": 4,
"summary": "The paper introduces TANGO, a framework for generating high-fidelity co-speech gesture videos using a motion graph-based retrieval approach. It addresses audio-mo... | {
"rating": "6;6;8;8",
"rating_avg": 7,
"confidence": "4;5;4;5",
"confidence_avg": 4.5,
"soundness": "3;3;4;4",
"soundness_avg": 3.5,
"contribution": "2;3;4;4",
"contribution_avg": 3.25,
"presentation": "2;3;3;4",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Oral",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.822423"
} | {
"id": "NhKMyMjEox",
"metareview": "The paper introduces a framework for generating high-fidelity co-speech gesture videos using a motion graph-based retrieval approach. It addresses audio-motion misalignment and visual artifacts by introducing two key innovations: a hierarchical audio-motion joint embedding space... | {
"decision": "Accept (Oral)"
} |
LbceJJc9h2 | 2410.05448v1 | Task Diversity Shortens the ICL Plateau | {
"content": "## Abstract\n\nAbstract In-context learning (ICL) describes a language model’s ability to generate outputs based on a set of input demonstrations and a subsequent query. To understand this remarkable capability, researchers have studied simplified, stylized models. These studies have consistently observ... | [
{
"id": "gtlEug7bap",
"initial_rating": 3,
"confidence": 4,
"soundness": 1,
"contribution": 1,
"presentation": 4,
"summary": "The paper explores the impact of multi-task learning (\"task diversity\" in the title) on the speed of training on individual tasks in the context of In-Context L... | {
"rating": "3;3;5;6",
"rating_avg": 4.25,
"confidence": "4;4;4;3",
"confidence_avg": 3.75,
"soundness": "2;1;2;3",
"soundness_avg": 2,
"contribution": "1;1;2;3",
"contribution_avg": 1.75,
"presentation": "3;4;4;2",
"presentation_avg": 3.25
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:01.823138"
} | {
"id": "0JwYf4nQ2J",
"metareview": "## Summary: \nThe paper investigates the effects of multi-task learning on the training speed of individual tasks when training models with In-Context Learning (ICL) tasks (synthetic tasks in the paper). Surprisingly, the results indicate that training on multiple diverse ICL ta... | {
"decision": "Reject"
} |
Lfy9q7Icp9 | 2410.03883v1 | DiSK: Differentially Private Optimizer with Simplified Kalman Filter for Noise Reduction | {
"content": "## Abstract\n\nAbstract Differential privacy (DP) offers a robust framework for safeguarding individual data privacy. To utilize DP in training modern machine learning models, differentially private optimizers have been widely used in recent years. A popular approach to privatize an optimizer is to clip... | [
{
"id": "Vj1sLt9YNU",
"initial_rating": 8,
"confidence": 3,
"soundness": 3,
"contribution": 4,
"presentation": 3,
"summary": "This paper introduces DiSK, a novel framework for enhancing differentially private (DP) optimizers by applying Kalman filtering to denoise privatized gradients, p... | {
"rating": "5;5;6;8",
"rating_avg": 6,
"confidence": "4;2;3;3",
"confidence_avg": 3,
"soundness": "3;3;3;3",
"soundness_avg": 3,
"contribution": "3;2;3;4",
"contribution_avg": 3,
"presentation": "4;2;2;3",
"presentation_avg": 2.75
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.824215"
} | {
"id": "4dgPlVmGnS",
"metareview": "This paper contributes to a growing body of work that aims to improve the performance (utility) of differentially private stochastic gradient descent (DP-SGD) by post-processing gradients. The main contribution is DiSK: a simplified Kalman Filter to denoise privatized gradients.... | {
"decision": "Accept (Poster)"
} |
LiUfN9h0Lx | 2406.18334v1 | Efficient and Accurate Explanation Estimation with Distribution Compression | {
"content": "## Abstract\n\nAbstract Exact computation of various machine learning explanations requires numerous model evaluations and in extreme cases becomes impractical. The computational cost of approximation increases with an ever-increasing size of data and model parameters. Many heuristics have been proposed... | [
{
"id": "DRefZ9biMH",
"initial_rating": 3,
"confidence": 5,
"soundness": 3,
"contribution": 3,
"presentation": 2,
"summary": "the paper introduces CTE uses distribution compression through kernel thinning to obtain a data sample to better and more efficiently approximate the marginal dis... | {
"rating": "3;8;8",
"rating_avg": 6.333333333333333,
"confidence": "5;3;2",
"confidence_avg": 3.3333333333333335,
"soundness": "2;3;3",
"soundness_avg": 2.6666666666666665,
"contribution": "1;3;3",
"contribution_avg": 2.3333333333333335,
"presentation": "2;3;3",
"presentation_avg": 2.66666666666666... | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Spotlight",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.825224"
} | {
"id": "nS5Nzd0RBA",
"metareview": "The paper introduces a compress-and-explain scheme to speedup attribution calculation. The reviewers are largely in agremeent that the proposed method is worthy of publication.",
"additional_comments": "No discussions needed because the reviewers are in agreement."
} | {
"decision": "Accept (Spotlight)"
} |
LikKyNlzgP | 2410.02601v1 | Diffusion & Adversarial Schrödinger Bridges via Iterative Proportional Markovian Fitting | {
"content": "## Abstract\n\nAbstract The Iterative Markovian Fitting (IMF) procedure based on iterative reciprocal and Markovian projections has recently been proposed as a powerful method for solving the Schrödinger Bridge problem. However, it has been observed that for the practical implementation of this procedur... | [
{
"id": "5H9QathFho",
"initial_rating": 6,
"confidence": 2,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "In the context of Schrödinger Bridge (SB) problems, the paper analyses theoretically some heuristic algorithms typically used in practice in the literature. Speci... | {
"rating": "3;6;6;6",
"rating_avg": 5.25,
"confidence": "4;3;3;2",
"confidence_avg": 3,
"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": "2026-01-14T22:16:01.826055"
} | {
"id": "Udlzq3fCMi",
"metareview": "(a) Summary of the paper's claims and findings:\n\nThe paper introduces the Iterative Proportional Markovian Fitting (IPMF) procedure, combining aspects of the Iterative Markovian Fitting (IMF) and Iterative Proportional Fitting (IPF) algorithms. The authors theoretically demons... | {
"decision": "Reject"
} |
LkzuPorQ5L | 2410.02506v1 | Cut the Crap: An Economical Communication Pipeline for LLM-based Multi-Agent Systems | {
"content": "## Abstract\n\nAbstract Recent advancements in large language model (LLM)-powered agents have shown that collective intelligence can significantly outperform individual capabilities, largely attributed to the meticulously designed inter-agent communication topologies. Though impressive in performance, e... | [
{
"id": "4K6REDPhNk",
"initial_rating": 6,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 2,
"summary": "The paper introduces AgentPrune, a communication framework designed for LLM-MA systems. AgentPrune addresses inefficiencies in traditional multi-agent communicati... | {
"rating": "3;5;6;6",
"rating_avg": 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": "2;3;2;2",
"presentation_avg": 2.25
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.827158"
} | {
"id": "Elp7d3JKdc",
"metareview": "This paper looks at optimizing the communication paradigms in multi-agent systems — roughly, “can we use fewer tokens over the course of a multi-agent task solve to achieve similar or better end results than a baseline without optimization.” Their focus is on pruning communicat... | {
"decision": "Accept (Poster)"
} |
Lnuy691O8Q | 2410.05248v1 | SFTMix: Elevating Language Model Instruction Tuning with Mixup Recipe | {
"content": "## Abstract\n\nAbstract To induce desired behaviors in large language models (LLMs) for interaction-driven tasks, the instruction-tuning stage typically trains LLMs on instruction-response pairs using the next-token prediction (NTP) loss.\nPrevious work aiming to improve instruction-tuning performance o... | [
{
"id": "9XMKiykEN6",
"initial_rating": 5,
"confidence": 4,
"soundness": 4,
"contribution": 3,
"presentation": 3,
"summary": "The paper introduces SFTMix, which uses Mixup-based regularization to interpolate the representations of training data with different confidence levels. This appr... | {
"rating": "3;3;3;5",
"rating_avg": 3.5,
"confidence": "4;4;4;4",
"confidence_avg": 4,
"soundness": "1;2;3;4",
"soundness_avg": 2.5,
"contribution": "1;2;2;3",
"contribution_avg": 2,
"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": "2026-01-14T22:16:01.828081"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
LoXJlAW3gU | 2403.03726v1 | Diffusion on language model encodings for protein sequence generation | {
"content": "## Abstract\n\nAbstract Protein design requires a deep understanding of the inherent complexities of the protein universe.\nWhile many efforts lean towards conditional generation or focus on specific families of proteins, the foundational task of unconditional generation remains underexplored and underv... | [
{
"id": "aB90uN6O8C",
"initial_rating": 5,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 2,
"summary": "This work presents an approach for unconditional protein sequence generation that operates via continuous diffusion in the latent space of protein sequence embedd... | {
"rating": "1;3;5;5;5;6",
"rating_avg": 4.166666666666667,
"confidence": "4;4;5;3;3;4",
"confidence_avg": 3.8333333333333335,
"soundness": "2;2;2;3;3;3",
"soundness_avg": 2.5,
"contribution": "2;2;2;3;3;3",
"contribution_avg": 2.5,
"presentation": "2;1;2;2;3;3",
"presentation_avg": 2.16666666666666... | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:01.828888"
} | {
"id": "49aHiRTU1E",
"metareview": "After all discussion phases have concluded, three reviewers judge the paper to be marginally above the acceptance threshold, and three reviewers judge the paper to be marginally or clearly below the acceptance threshold. \n\nAmong other things, as positives the reviewers highlig... | {
"decision": "Reject"
} |
Lp40Z40N07 | 2410.18978v2 | Framer: Interactive Frame Interpolation | {
"content": "## Abstract\n\nAbstract We propose Framer for interactive frame interpolation, which targets producing smoothly transitioning frames between two images as per user creativity.\nConcretely, besides taking the start and end frames as inputs, our approach supports customizing the transition process by tail... | [
{
"id": "MwTwKdGOr3",
"initial_rating": 5,
"confidence": 3,
"soundness": 3,
"contribution": 2,
"presentation": 2,
"summary": "The paper proposes a new technique for generating frame interpolations that can incorporate user input to drive the interpolation process between two given frames... | {
"rating": "5;5;6;8",
"rating_avg": 6,
"confidence": "5;3;3;4",
"confidence_avg": 3.75,
"soundness": "3;3;3;4",
"soundness_avg": 3.25,
"contribution": "2;2;2;4",
"contribution_avg": 2.5,
"presentation": "4;2;3;4",
"presentation_avg": 3.25
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.829734"
} | {
"id": "2pOA8fMvDo",
"metareview": "The paper introduces Framer, an method for using user-provided trajectories as guidance in frame interpolation or morphing. Reviewers praised the paper's clear motivation and impressive performance, but some questioned its practical applicability and technical innovation. Revie... | {
"decision": "Accept (Poster)"
} |
LphpWGimIa | 2406.17759v1 | Interpreting Attention Layer Outputs with Sparse Autoencoders | {
"content": "## Abstract\n\nAbstract Decomposing model activations into interpretable components is a key open problem in mechanistic interpretability. Sparse autoencoders (SAEs) are a popular method for decomposing the internal activations of trained transformers into sparse, interpretable features, and have been a... | [
{
"id": "qWWsMeiY6D",
"initial_rating": 6,
"confidence": 3,
"soundness": 2,
"contribution": 3,
"presentation": 3,
"summary": "This paper applies sparse autoencoders (SAEs), a popular technique in mechanistic interpretability, to the output of attention layers in pre-trained Transformer m... | {
"rating": "3;6;6",
"rating_avg": 5,
"confidence": "4;3;3",
"confidence_avg": 3.3333333333333335,
"soundness": "1;3;2",
"soundness_avg": 2,
"contribution": "1;3;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": "2026-01-14T22:16:01.830624"
} | {
"id": "vZNoVU4hnL",
"metareview": "The paper introduces a method to apply sparse autoencoders to interpret attention layer outputs in transformers. \n\nDespite some reviewers acknowledging the potential interest for the interpretability community, the main reservations remained: the ad hoc interpretability metric... | {
"decision": "Reject"
} |
LqBPkN8GTH | 2406.12805v2 | AITTI: Learning Adaptive Inclusive Token for Text-to-Image Generation | {
"content": "## Abstract\n\nAbstract Despite the high-quality results of text-to-image generation, stereotypical biases have been spotted in their generated contents, compromising the fairness of generative models.\nIn this work, we propose to learn adaptive inclusive tokens to shift the attribute distribution of th... | [
{
"id": "k30nngyggw",
"initial_rating": 5,
"confidence": 4,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "The paper focuses on addressing the bias problem in current T2I models. To tackle this, the authors introduce a prompt-tuning approach based on an adaptive mappin... | {
"rating": "3;5;5;5",
"rating_avg": 4.5,
"confidence": "3;4;4;4",
"confidence_avg": 3.75,
"soundness": "3;2;3;3",
"soundness_avg": 2.75,
"contribution": "2;2;2;2",
"contribution_avg": 2,
"presentation": "2;2;2;3",
"presentation_avg": 2.25
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:01.831397"
} | {
"id": "NYQME16cVx",
"metareview": "This paper introduces AITTI, a method using adaptive inclusive tokens to mitigate biases in text-to-image generation. The approach involves tuning an adaptive mapping network with inclusive samples using an anchor loss to mitigate biases without requiring explicit attribute spec... | {
"decision": "Reject"
} |
LqTz13JS2P | 2402.09721v4 | Generalized Principal-Agent Problem with a Learning Agent | {
"content": "## Abstract\n\nAbstract Classic principal-agent problems such as Stackelberg games, contract design, and Bayesian persuasion, often assume that the agent is able to best respond to the principal’s committed strategy.\nWe study repeated generalized principal-agent problems under the assumption that the p... | [
{
"id": "zDCpipfo7C",
"initial_rating": 6,
"confidence": 2,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "The paper considers the principal's adaptive strategy design problem against a non-regret learning agent. The paper shows that the problem can be reduced from aga... | {
"rating": "5;6;6;8",
"rating_avg": 6.25,
"confidence": "3;4;2;4",
"confidence_avg": 3.25,
"soundness": "3;3;3;4",
"soundness_avg": 3.25,
"contribution": "2;3;2;4",
"contribution_avg": 2.75,
"presentation": "2;3;3;4",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Spotlight",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.832304"
} | {
"id": "j7Uj8MAniF",
"metareview": "The paper studies (generalized) principal-agent problems in EconCS (including Bayesian persuasion and contract design) using a no-regret framework. I agree with the authors that the \"clean reduction and the generality of the analysis\" (which applies to all principal-agent prob... | {
"decision": "Accept (Spotlight)"
} |
Lr8IIc1rB8 | 2410.03132v2 | Autoregressive Action Sequence Learning for Robotic Manipulation | {
"content": "## Abstract\n\nAbstract Autoregressive models have demonstrated remarkable success in natural language processing. In this work, we design a simple yet effective autoregressive architecture for robotic manipulation tasks. We propose the Chunking Causal Transformer (CCT), which extends the next-single-to... | [
{
"id": "bELzd6VFEA",
"initial_rating": 3,
"confidence": 4,
"soundness": 2,
"contribution": 2,
"presentation": 3,
"summary": "The paper proposes an autoregressive architecture for policy learning. The model architecture, called chunking causal transformer, predicts several future actions... | {
"rating": "3;3;5;5",
"rating_avg": 4,
"confidence": "4;4;3;4",
"confidence_avg": 3.75,
"soundness": "2;2;2;2",
"soundness_avg": 2,
"contribution": "2;2;1;2",
"contribution_avg": 1.75,
"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": "2026-01-14T22:16:01.833336"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
LtBD5fFHB7 | 2403.20193v2 | Motion Inversion for Video Customization | {
"content": "## Abstract\n\nAbstract In this work, we present a novel approach for motion customization in video generation, addressing the widespread gap in the exploration of motion representation within video generative models. Recognizing the unique challenges posed by the spatiotemporal nature of video, our met... | [
{
"id": "yqvye2D2AW",
"initial_rating": 6,
"confidence": 5,
"soundness": 3,
"contribution": 3,
"presentation": 4,
"summary": "The paper proposes a motion customization method for video diffusion models. Specifically, it introduces two types of motion embeddings to capture the global rela... | {
"rating": "3;5;6;6",
"rating_avg": 5,
"confidence": "4;5;4;5",
"confidence_avg": 4.5,
"soundness": "3;2;3;3",
"soundness_avg": 2.75,
"contribution": "3;1;3;3",
"contribution_avg": 2.5,
"presentation": "3;3;3;4",
"presentation_avg": 3.25
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:01.833951"
} | {
"id": "KtaOD70sEy",
"metareview": "The paper addresses the problem of motion customization for video models (given a video, generate a video with its approximate motion) and proposes a \"motion embedding\" which is designed and trained specifically to capture the motion in the source video. Specifically, this mot... | {
"decision": "Reject"
} |
Lut5t3qElA | 2407.03824v1 | Unsupervised Disentanglement of Content and Style via Variance-Invariance Constraints | {
"content": "## Abstract\n\nAbstract We contribute an unsupervised method that effectively learns from raw observation and disentangles its latent space into content and style representations. Unlike most disentanglement algorithms that rely on domain-specific labels and knowledge, our method is based on the insight... | [
{
"id": "uKzrG5lJAq",
"initial_rating": 6,
"confidence": 2,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "They propose an unsupervised method to disentangle content and style representations. They use vector-quantized autoencoder architecture and incorporate variance... | {
"rating": "3;5;5;6;8",
"rating_avg": 5.4,
"confidence": "4;2;4;2;5",
"confidence_avg": 3.4,
"soundness": "3;3;2;3;3",
"soundness_avg": 2.8,
"contribution": "2;2;2;3;3",
"contribution_avg": 2.4,
"presentation": "3;3;3;3;4",
"presentation_avg": 3.2
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.834554"
} | {
"id": "qjZUXtqJww",
"metareview": "This paper introduces V3 (variance-versus-invariance), an unsupervised method designed to effectively disentangle content and style representations from sequences of observations without relying on domain-specific labels or knowledge.\n\nThe reviewers acknowledged the novelty an... | {
"decision": "Accept (Poster)"
} |
LuytzzohTa | 2408.07199v1 | Agent Q: Advanced Reasoning and Learning for Autonomous AI Agents | {
"content": "## Abstract\n\nAbstract Large Language Models (LLMs) have shown remarkable capabilities in natural language tasks requiring complex reasoning, yet their application in agentic, multi-step reasoning within interactive environments remains a difficult challenge. Traditional supervised pre-training on stat... | [
{
"id": "oSxFwCbYaR",
"initial_rating": 5,
"confidence": 4,
"soundness": 2,
"contribution": 2,
"presentation": 1,
"summary": "The paper proposes Agent Q, an online fine-tuning framework for enhancing reasoning capabilities in autonomous AI agents for web navigation tasks. Agent Q combine... | {
"rating": "5;5;5;5;8",
"rating_avg": 5.6,
"confidence": "4;5;4;4;3",
"confidence_avg": 4,
"soundness": "2;4;2;2;3",
"soundness_avg": 2.6,
"contribution": "3;2;1;2;3",
"contribution_avg": 2.2,
"presentation": "1;2;3;1;4",
"presentation_avg": 2.2
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:01.835236"
} | {
"id": "flFNt4WANG",
"metareview": "This paper proposes an online framework to improve agents' autonomous reasoning capabilities in web navigation. There are two main concerns. First, the method and experiments (including ablation studies) lack details, which make it hard to follow the manuscript. Second, the nove... | {
"decision": "Reject"
} |
LvDwwAgMEW | 2310.11589v1 | Eliciting Human Preferences with Language Models | {
"content": "## Abstract\n\nAbstract Language models (LMs) can be directed to perform target tasks by using labeled examples or natural language prompts.\nBut selecting examples or writing prompts for can be challenging—especially in tasks that involve unusual edge cases, demand precise articulation of nebulous pref... | [
{
"id": "macLSVrMlB",
"initial_rating": 8,
"confidence": 5,
"soundness": 4,
"contribution": 3,
"presentation": 4,
"summary": "The paper contributes the GATE framework, an alternative to tradition methods of gathering human preferences for various means including alignment, which is used ... | {
"rating": "5;5;8;8",
"rating_avg": 6.5,
"confidence": "3;4;5;5",
"confidence_avg": 4.25,
"soundness": "2;2;4;4",
"soundness_avg": 3,
"contribution": "2;3;3;3",
"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": "2026-01-14T22:16:01.836142"
} | {
"id": "LaJl7x2SiA",
"metareview": "Overall this was seen as a good paper, and an important contribution to the community. The paper is clear and well written. It studies three tasks that are close to real world tasks. The key benefit that this paper provides is a method to actively elicit user preferences rath... | {
"decision": "Accept (Poster)"
} |
LvNROciCne | 2410.17881v1 | AdaRankGrad: Adaptive Gradient Rank and Moments for Memory-Efficient LLMs Training and Fine-Tuning | {
"content": "## Abstract\n\nAbstract Training and fine-tuning large language models (LLMs) come with challenges related to memory and computational requirements due to the increasing size of the model weights and the optimizer states. Various techniques have been developed to tackle these challenges, such as low-ran... | [
{
"id": "MBWHwIPeEa",
"initial_rating": 8,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "The authors propose AdaRankGrad that learns to adaptively perform low-rank gradient updates in order to reduce the memory requirements needed to train a large lan... | {
"rating": "5;5;6;6;8",
"rating_avg": 6,
"confidence": "3;3;3;3;4",
"confidence_avg": 3.2,
"soundness": "3;3;3;3;3",
"soundness_avg": 3,
"contribution": "3;3;2;3;3",
"contribution_avg": 2.8,
"presentation": "4;3;3;3;3",
"presentation_avg": 3.2
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.837215"
} | {
"id": "d2zokF70bB",
"metareview": "In this paper, the authors proposed a memory-efficient low-rank gradient update algorithm for LLMs training and fine-tuning.\n\nThe reviewers raised some concerns and questions regarding the empirical evaluation of the proposed algorithm. Most of the concerns are addressed durin... | {
"decision": "Accept (Poster)"
} |
LvRQgsvd5V | 2411.07007v1 | Non-Adversarial Inverse Reinforcement Learning via Successor Feature Matching | {
"content": "## Abstract\n\nAbstract In inverse reinforcement learning (IRL), an agent seeks to replicate expert demonstrations through interactions with the environment.\nTraditionally, IRL is treated as an adversarial game, where an adversary searches over reward models, and a learner optimizes the reward through ... | [
{
"id": "BLsT4riSIU",
"initial_rating": 6,
"confidence": 3,
"soundness": 2,
"contribution": 3,
"presentation": 3,
"summary": "This work considers inverse RL in the state-only setting. Leveraging the linear structure of returns, as inner product of successor features and reward weight, th... | {
"rating": "3;6;6;8",
"rating_avg": 5.75,
"confidence": "3;3;3;3",
"confidence_avg": 3,
"soundness": "3;3;2;3",
"soundness_avg": 2.75,
"contribution": "2;3;3;4",
"contribution_avg": 3,
"presentation": "2;3;3;3",
"presentation_avg": 2.75
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.838006"
} | {
"id": "DCuUnIwA1n",
"metareview": "This paper studies Inverse Reinforcement Learning (IRL) via a successor-feature-based approach, the Successor Feature Matching (SFM), which is non-adversarial and does not require expert action labels. In particular, SFM directly optimizes a policy to match the expert's success... | {
"decision": "Accept (Poster)"
} |
LvuSFvGShf | 2410.01866v1 | House of Cards: Massive Weights in LLMs | {
"content": "## Abstract\n\nAbstract Massive activations, which manifest in specific feature dimensions of hidden states, introduce a significant bias in large language models (LLMs), leading to an overemphasis on the corresponding token. In this paper, we identify that massive activations originate not from the hid... | [
{
"id": "6ZoIVEbe5J",
"initial_rating": 5,
"confidence": 4,
"soundness": 2,
"contribution": 2,
"presentation": 3,
"summary": "This paper reveals that large language models (LLMs) have a bias towards specific tokens due to massive activations in their hidden states, which stem from the in... | {
"rating": "3;5;5;8",
"rating_avg": 5.25,
"confidence": "4;4;4;4",
"confidence_avg": 4,
"soundness": "1;2;2;3",
"soundness_avg": 2,
"contribution": "2;2;2;3",
"contribution_avg": 2.25,
"presentation": "3;3;3;4",
"presentation_avg": 3.25
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:01.838741"
} | {
"id": "G0BfEUoFiR",
"metareview": "(a) Summary\n\nThe paper discovers massive weights as causes for massive activations, and proposes MacDrop, a dropout-based method targeting massive weights in LLMs, to improve fine-tuning efficiency and performance.\n\n(b) Strengths\n\nIdentifies massive weights as causes for r... | {
"decision": "Reject"
} |
LwAG269lIq | 2401.17177v3 | Data-Driven Discovery of PDEs via the Adjoint Method | {
"content": "## Abstract\n\nAbstract In this work, we present an adjoint-based method for discovering the underlying governing partial differential equations (PDEs) given data. The idea is to consider a parameterized PDE in a general form, and formulate the optimization problem that minimizes the error of PDE soluti... | [
{
"id": "qvc0sq5Iop",
"initial_rating": 3,
"confidence": 4,
"soundness": 2,
"contribution": 2,
"presentation": 2,
"summary": "The authors claim that they can discover some PDE via an adjoint-based method.",
"strengths": "The topic looks very interesting.",
"weaknesses": "The manu... | {
"rating": "3;3;3;3",
"rating_avg": 3,
"confidence": "3;5;5;4",
"confidence_avg": 4.25,
"soundness": "3;2;1;2",
"soundness_avg": 2,
"contribution": "2;2;2;2",
"contribution_avg": 2,
"presentation": "3;1;2;2",
"presentation_avg": 2
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:01.839414"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
Ly0SQh7Urv | 2410.01606v1 | Automated Red Teaming with GOAT: the Generative Offensive Agent Tester | {
"content": "## Abstract\n\nAbstract Red teaming assesses how large language models (LLMs) can produce content that violates norms, policies, and rules set during their safety training.\nHowever, most existing automated methods in the literature are not representative of the way humans tend to interact with AI model... | [
{
"id": "fQEi6Nmo3T",
"initial_rating": 6,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "This work offers a novel way to simulate and automate how human users may use publicly available jailbreak methods to circumvent the safety training of LMs in lon... | {
"rating": "5;5;5;6;6",
"rating_avg": 5.4,
"confidence": "4;4;4;3;4",
"confidence_avg": 3.8,
"soundness": "2;3;3;3;3",
"soundness_avg": 2.8,
"contribution": "2;2;3;3;3",
"contribution_avg": 2.6,
"presentation": "3;3;2;3;3",
"presentation_avg": 2.8
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:01.840040"
} | {
"id": "eMifweBpPD",
"metareview": "The reviewers for this paper have diverse assessments, with overall ratings for the paper being borderline. On the positive side, the reviewers acknowledged that the proposed automated red-teaming method leverages an interesting idea of multi-turn conservations simulating human ... | {
"decision": "Reject"
} |
LyJi5ugyJx | 2410.11081v1 | Simplifying, Stabilizing and Scaling Continuous-time Consistency Models | {
"content": "## Abstract\n\nAbstract Consistency models (CMs) are a powerful class of diffusion-based generative models optimized for fast sampling. Most existing CMs are trained using discretized timesteps, which introduce additional hyperparameters and are prone to discretization errors. While continuous-time form... | [
{
"id": "0qzpk8E8Ds",
"initial_rating": 8,
"confidence": 5,
"soundness": 4,
"contribution": 4,
"presentation": 3,
"summary": "This paper investigates a fundamental topic in consistency models (CMs), specifically the challenges of discretization errors and the resulting training stability... | {
"rating": "6;6;8;8;10",
"rating_avg": 7.6,
"confidence": "4;3;4;5;4",
"confidence_avg": 4,
"soundness": "3;3;3;4;4",
"soundness_avg": 3.4,
"contribution": "3;3;4;4;4",
"contribution_avg": 3.6,
"presentation": "3;4;3;3;4",
"presentation_avg": 3.4
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Oral",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.840856"
} | {
"id": "oYqhEjmNaX",
"metareview": "Consistency models (CMs) are specializations of diffusion models for the purpose of faster sampling. While existing CMs are faster, they are difficult to train, and the quality lags behind other diffusion models. This work provides a close study and several improvements that the... | {
"decision": "Accept (Oral)"
} |
M23dTGWCZy | 2409.04109v1 | Can LLMs Generate Novel Research Ideas? A Large-Scale Human Study with 100+ NLP Researchers | {
"content": "## Abstract\n\nAbstract Recent advancements in large language models (LLMs) have sparked optimism about their potential to accelerate scientific discovery, with a growing number of works proposing research agents that autonomously generate and validate new ideas.\nDespite this, no evaluations have shown... | [
{
"id": "05SiJAQyBq",
"initial_rating": 6,
"confidence": 4,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "This paper evaluates if LLM based RAG agents are capable of generating novel research ideas for prompting techniques by comparing them with expert researchers. Th... | {
"rating": "5;6;6;6",
"rating_avg": 5.75,
"confidence": "3;3;4;4",
"confidence_avg": 3.5,
"soundness": "2;3;4;3",
"soundness_avg": 3,
"contribution": "3;3;2;2",
"contribution_avg": 2.5,
"presentation": "3;3;4;3",
"presentation_avg": 3.25
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.842256"
} | {
"id": "oLJ8bE4OB8",
"metareview": "**Summary**\n\nThis paper aims to study whether LLMs can be scientifically creative and then evaluates if LLM based RAG agents are capable of generating novel research ideas for prompting techniques by comparing them with expert researchers. \nThis paper addresses the question t... | {
"decision": "Accept (Poster)"
} |
M4fhjfGAsZ | 2410.01727v1 | Automated Knowledge Concept Annotation and Question Representation Learning for Knowledge Tracing | {
"content": "## Abstract\n\nAbstract Knowledge tracing (KT) is a popular approach for modeling students’ learning progress over time, which can enable more personalized and adaptive learning. However, existing KT approaches face two major limitations: (1) they rely heavily on expert-defined knowledge concepts (KCs) ... | [
{
"id": "u4OoQWwiZU",
"initial_rating": 5,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "The authors address two primary shortcomings of existing work on knowledge tracing (KT): (1) mapping questions to knowledge concepts (KCs) being evaluated (I beli... | {
"rating": "3;5;6",
"rating_avg": 4.666666666666667,
"confidence": "5;4;2",
"confidence_avg": 3.6666666666666665,
"soundness": "1;3;3",
"soundness_avg": 2.3333333333333335,
"contribution": "1;3;3",
"contribution_avg": 2.3333333333333335,
"presentation": "3;3;3",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:01.843318"
} | {
"id": "8RRKOEFP1q",
"metareview": "Summary:\n\nThis paper presents KCQRL, a framework for automated knowledge concept (KC) annotation and question representation learning to improve the effectiveness of any existing knowledge tracing (KT) model. It proposes an automated KC annotation process using large language ... | {
"decision": "Reject"
} |
M5LGyR71yS | 2409.08239v1 | Source2Synth: Synthetic Data Generation and Curation Grounded in Real Data Sources | {
"content": "## Abstract\n\nAbstract Large Language Models still struggle in challenging scenarios that leverage structured data, complex reasoning, or tool usage. In this paper, we propose Source2Synth : a new method that can be used for teaching LLMs new skills without\nrelying on costly human annotations. Source2... | [
{
"id": "cBrvpkVQao",
"initial_rating": 6,
"confidence": 4,
"soundness": 3,
"contribution": 4,
"presentation": 1,
"summary": "This works presents an end-to-end pipeline to generate synthetic training data based on some data source. The pipeline consists of three steps, namely data genera... | {
"rating": "3;3;5;6",
"rating_avg": 4.25,
"confidence": "4;4;3;4",
"confidence_avg": 3.75,
"soundness": "2;2;2;3",
"soundness_avg": 2.25,
"contribution": "2;3;2;4",
"contribution_avg": 2.75,
"presentation": "1;2;3;1",
"presentation_avg": 1.75
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:01.844228"
} | {
"id": "JfhTxWvuq8",
"metareview": "This paper presents Source2Synth, a data augmentation method for improving LLM performance in low-data regimes without human annotations. The key strengths are the novel self-augmentation approach and strong empirical results showing 20-25% improvements on WikiSQL and HotpotQA b... | {
"decision": "Reject"
} |
M5t0WvjfCg | 2403.14614v1 | AdaIR: Adaptive All-in-One Image Restoration via Frequency Mining and Modulation | {
"content": "## Abstract\n\nAbstract In the image acquisition process, various forms of degradation, including noise, blur, haze, and rain, are frequently introduced. These degradations typically arise from the inherent limitations of cameras or unfavorable ambient conditions. To recover clean images from their degr... | [
{
"id": "QsqGWjL0lH",
"initial_rating": 6,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "The paper presents AdaIR, which aims to address multiple forms of degradation, such as noise, blur, haze, and rain, within a single model. AdaIR is motivated by t... | {
"rating": "3;5;6;8",
"rating_avg": 5.5,
"confidence": "4;5;4;5",
"confidence_avg": 4.5,
"soundness": "2;2;3;4",
"soundness_avg": 2.75,
"contribution": "2;2;3;3",
"contribution_avg": 2.5,
"presentation": "3;2;3;3",
"presentation_avg": 2.75
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.844931"
} | {
"id": "fpWWexE8xB",
"metareview": "All reviewers agree on the innovation and novelty of the proposed work. Reviewers had individual questions regarding the presentation of the results, which the authors responded rigorously. This led to reviewers increasing their initial score after the rebuttal to the average sc... | {
"decision": "Accept (Poster)"
} |
M5u38Os65F | 2310.04614v3 | Variance Reduced Distributed Non-Convex Optimization Using Matrix Stepsizes | {
"content": "## Abstract\n\nAbstract Matrix-stepsized gradient descent algorithms have been shown to have superior performance in non-convex optimization problems compared to their scalar counterparts.\nThe det-CGD algorithm, as introduced by Li et al. ( 2024b ) , leverages matrix stepsizes to perform compressed gra... | [
{
"id": "3w5mNR5J4A",
"initial_rating": 5,
"confidence": 3,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "This paper provided two novel federated learning optimization methods based on variance-reduced matrix stepsizes. The theoretical analysis shows the communication... | {
"rating": "3;5;5;5",
"rating_avg": 4.5,
"confidence": "3;3;3;3",
"confidence_avg": 3,
"soundness": "3;3;3;3",
"soundness_avg": 3,
"contribution": "2;2;2;2",
"contribution_avg": 2,
"presentation": "2;3;2;3",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:01.846321"
} | {
"id": "Smowu7xTuX",
"metareview": "This paper extends the det-CGD algorithm, which uses matrix stepsizes for non-convex optimization, by introducing two variance-reduced versions: det-MARINA and det-DASHA. These methods improve iteration and communication complexities over the existing methods, with both theoreti... | {
"decision": "Reject"
} |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.