paper_id string | arxiv_id string | title string | markdown dict | reviews list | scores dict | metadata dict | meta_review dict | decision dict |
|---|---|---|---|---|---|---|---|---|
qdbluGtEpL | 2410.14966v1 | Attack as Defense: Run-time Backdoor Implantation for Image Content Protection | {
"content": "## Abstract\n\nAbstract As generative models achieve great success, tampering and modifying the sensitive image contents (i.e., human faces, artist signatures, commercial logos, etc.) have induced a significant threat with social impact.\nThe backdoor attack is a method that implants vulnerabilities in ... | [
{
"id": "tEcC2Jf0TE",
"initial_rating": 5,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "This paper presents a method for image content protection by implanting run-time backdoors in image-editing models and triggering the backdoors when sensitive con... | {
"rating": "5;5;5;6;8",
"rating_avg": 5.8,
"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": "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": "2026-01-14T22:16:03.719838"
} | {
"id": "RszIHzHTun",
"metareview": "In this paper, the authors proposed a run-time backdoor attack to prevent the sensitive region from being tampered with.\nThere are several comments raised by the reviewers, so the authors replied to 18 questions raised by the reviewers, provided additional 7 experiments, and in... | {
"decision": "Reject"
} |
qeXcMutEZY | 2403.08728v1 | Ambient Diffusion Posterior Sampling: Solving Inverse Problems with Diffusion Models Trained on Corrupted Data | {
"content": "## Abstract\n\nAbstract We provide a framework for solving inverse problems with diffusion models learned from linearly corrupted data. Our method, Ambient Diffusion Posterior Sampling (A-DPS), leverages a generative model pre-trained on one type of corruption (e.g. image inpainting) to perform posterio... | [
{
"id": "w2GAyLE4GY",
"initial_rating": 6,
"confidence": 4,
"soundness": 2,
"contribution": 3,
"presentation": 4,
"summary": "This paper uses ambient diffusion models as a prior for solving ill-posed inverse problems. The ambient diffusion is trained on corrupted data acquired from a lin... | {
"rating": "5;5;6;6;8",
"rating_avg": 6,
"confidence": "5;4;4;5;4",
"confidence_avg": 4.4,
"soundness": "3;2;2;2;4",
"soundness_avg": 2.6,
"contribution": "3;1;3;3;4",
"contribution_avg": 2.8,
"presentation": "3;2;2;3;3",
"presentation_avg": 2.6
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.720500"
} | {
"id": "M1MOrGlOTj",
"metareview": "This paper combines the recently proposed ambient diffusion with diffusion posterior sampling (DPS) to make ambient diffusion posterior sampling (A-DPS). They adapt ambient diffusion to allow training on corrupted (say subsampled) Fourier measurements. This results in a self-sup... | {
"decision": "Accept (Poster)"
} |
qeYa5LRveW | 2410.09141v1 | ACER: Automatic Language Model Context Extension via Retrieval | {
"content": "## Abstract\n\nAbstract Long-context modeling is one of the critical capabilities of language AI for digesting and reasoning over complex information pieces. In practice, long-context capabilities are typically built into a pre-trained language model (LM) through a carefully designed context extension s... | [
{
"id": "bzDDkBRtog",
"initial_rating": 5,
"confidence": 3,
"soundness": 2,
"contribution": 3,
"presentation": 2,
"summary": "This paper explores the challenge of building language models with strong long-context processing capabilities, which are essential for tasks that require digesti... | {
"rating": "3;5;5;5",
"rating_avg": 4.5,
"confidence": "5;4;3;3",
"confidence_avg": 3.75,
"soundness": "3;3;2;2",
"soundness_avg": 2.5,
"contribution": "2;3;2;3",
"contribution_avg": 2.5,
"presentation": "3;2;3;2",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:03.721328"
} | {
"id": "iViUey1Oxb",
"metareview": "Long-context language modeling remains challenging for open-weight models. This paper presents a new approach for long-context LM that leverages retrieval and data synthesis. This method first uses retrieval to identify a small number of relevant documents, generates CoT answers... | {
"decision": "Reject"
} |
qezVbskHmi | 2405.11828v1 | Federated Learning for Time-Series Healthcare Sensing with Incomplete Modalities | {
"content": "## Abstract\n\nAbstract. Many mobile sensing applications utilize data from various modalities, including motion and physiological sensors in mobile and wearable devices. Federated Learning (FL) is particularly suitable for these applications thanks to its privacy-preserving feature. However, challenges... | [
{
"id": "iDmEMXVHdU",
"initial_rating": 6,
"confidence": 3,
"soundness": 3,
"contribution": 4,
"presentation": 4,
"summary": "This paper tries to develop an efficient and scalable federated learning algorithms that handles incomplete modalities while maintaining high accuracy. Healthcare... | {
"rating": "3;3;5;6",
"rating_avg": 4.25,
"confidence": "5;4;4;3",
"confidence_avg": 4,
"soundness": "2;3;2;3",
"soundness_avg": 2.5,
"contribution": "2;2;2;4",
"contribution_avg": 2.5,
"presentation": "3;2;3;4",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:03.722163"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
qfU5S4cddQ | 2408.05215v1 | Physics-Informed Weakly Supervised Learning for Interatomic Potentials | {
"content": "## Abstract\n\nAbstract Machine learning plays an increasingly important role in computational chemistry and materials science, complementing computationally intensive ab initio and first-principles methods. Despite their utility, machine-learning models often lack generalization capability and robustne... | [
{
"id": "wPMwNsNijX",
"initial_rating": 3,
"confidence": 4,
"soundness": 2,
"contribution": 2,
"presentation": 2,
"summary": "The authors propose two self-supervised losses to train machine learning interatomic potentials. One is based on a Taylor-expansion approach, and the other is bas... | {
"rating": "3;3;6;6",
"rating_avg": 4.5,
"confidence": "5;4;5;3",
"confidence_avg": 4.25,
"soundness": "2;2;4;3",
"soundness_avg": 2.75,
"contribution": "2;2;3;3",
"contribution_avg": 2.5,
"presentation": "4;2;3;3",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:03.722906"
} | {
"id": "6vzs2lwOQu",
"metareview": "The submission presents a physics-informed weakly supervised method to improve the generalization of machine-learning force field models (MLFFs), especially in situations where labeled data are scarce. Two losses are proposed based on the physics fact that the force is the negat... | {
"decision": "Reject"
} |
qgsXsqahMq | 2410.09570v1 | GETS: Ensemble Temperature Scaling for Calibration in Graph Neural Networks | {
"content": "## Abstract\n\nAbstract Graph Neural Networks (GNNs) deliver strong classification results but often suffer from poor calibration performance, leading to overconfidence or underconfidence.\nThis is particularly problematic in high-stakes applications where accurate uncertainty estimates are essential.\n... | [
{
"id": "rQ92P4589K",
"initial_rating": 8,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "The paper proposes graph ensemble temperature scaling (GETS), a novel calibration technique for node-classification GNNs.\nGETS is a parametric calibration method... | {
"rating": "6;6;6;8",
"rating_avg": 6.5,
"confidence": "3;4;4;4",
"confidence_avg": 3.75,
"soundness": "3;2;3;3",
"soundness_avg": 2.75,
"contribution": "2;2;2;3",
"contribution_avg": 2.25,
"presentation": "4;2;3;3",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Spotlight",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.723663"
} | {
"id": "suyeskUZxC",
"metareview": "The paper presents a novel approach, Graph Ensemble Temperature Scaling (GETS), aimed at improving the calibration of Graph Neural Networks (GNNs). By leveraging various input sources and employing a Mixture-of-Experts architecture, GETS effectively reduces expected calibration ... | {
"decision": "Accept (Spotlight)"
} |
qh1goDZ0ZQ | 2406.02500v2 | Towards Efficient Mixture of Experts: A Holistic Study of Compression Techniques | {
"content": "## Abstract\n\nAbstract Scaling large language models has revolutionized the performance across diverse domains, yet the continual growth in model size poses significant challenges for real-world deployment. The Mixture of Experts (MoE) approach addresses this by dynamically selecting and activating onl... | [
{
"id": "ZntN6M9Avp",
"initial_rating": 5,
"confidence": 3,
"soundness": 3,
"contribution": 1,
"presentation": 3,
"summary": "This paper presents a comprehensive study on compression techniques for Mixture of Experts (MoE) architectures to enhance efficiency and scalability in large lang... | {
"rating": "3;5;5",
"rating_avg": 4.333333333333333,
"confidence": "4;4;3",
"confidence_avg": 3.6666666666666665,
"soundness": "3;2;3",
"soundness_avg": 2.6666666666666665,
"contribution": "2;2;1",
"contribution_avg": 1.6666666666666667,
"presentation": "3;3;3",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:03.724349"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
qn9tBYQHGi | 2403.16843v3 | Do LLM Agents Have Regret? A Case Study in Online Learning and Games | {
"content": "## Abstract\n\nAbstract Large language models (LLMs) have been increasingly employed for (interactive) decision-making, via the development of LLM-based autonomous agents. Despite their emerging successes, the performance of LLM agents in decision-making has not been fully investigated through quantitat... | [
{
"id": "SxSqqWT1eR",
"initial_rating": 6,
"confidence": 2,
"soundness": 2,
"contribution": 3,
"presentation": 3,
"summary": "This paper studies large language models from a game-theoretic perspective. Observing that LLM-agents are becoming prevalent, they investigate the performance of ... | {
"rating": "5;5;6;6",
"rating_avg": 5.5,
"confidence": "2;3;4;2",
"confidence_avg": 2.75,
"soundness": "3;4;4;2",
"soundness_avg": 3.25,
"contribution": "2;3;3;3",
"contribution_avg": 2.75,
"presentation": "2;2;3;3",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.727107"
} | {
"id": "VvpUiUj3Ux",
"metareview": "Summary:\nThis work explores the use of large language model (LLM) agents in playing games. The authors conduct experiments across several LLM instances and problem settings, delivering key insights into LLM performance in gaming scenarios, such as demonstrating that most LLM ag... | {
"decision": "Accept (Poster)"
} |
qnAZqlMGTB | 2411.03628v1 | StreamingBench: Assessing the Gap for MLLMs to Achieve Streaming Video Understanding | {
"content": "## Abstract\n\nAbstract The rapid development of Multimodal Large Language Models (MLLMs) has expanded their capabilities from image comprehension to video understanding. However, most of these MLLMs focus primarily on offline video comprehension, necessitating extensive processing of all video frames b... | [
{
"id": "z3AD6FocQ6",
"initial_rating": 3,
"confidence": 4,
"soundness": 2,
"contribution": 2,
"presentation": 3,
"summary": "This work proposes a benchmark called StreamingBench to evaluate video LLM capabilities in streaming settings. StreamingBench introduces several tasks tailored to... | {
"rating": "3;6;6;8",
"rating_avg": 5.75,
"confidence": "4;5;5;4",
"confidence_avg": 4.5,
"soundness": "2;4;3;4",
"soundness_avg": 3.25,
"contribution": "2;3;3;3",
"contribution_avg": 2.75,
"presentation": "3;3;3;4",
"presentation_avg": 3.25
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:03.728943"
} | {
"id": "J4Ui5jRylC",
"metareview": "The paper presented a benchmark for stream video understanding. The paper received mixed ratings from four reviewers. Although some of the reviewers appreciated the importance of stream video understanding and the creation of a corresponding benchmark to advance this research ar... | {
"decision": "Reject"
} |
qnGir4dyu9 | 2405.18406v3 | RACCooN: A Versatile Instructional Video Editing Framework with Auto-Generated Narratives | {
"content": "## Abstract\n\nAbstract Recent video generative models primarily rely on carefully written text prompts for specific tasks, like inpainting or style editing. They require labor-intensive textual descriptions for input videos, hindering their flexibility to adapt personal/raw videos to user specification... | [
{
"id": "kRQXKsgREg",
"initial_rating": 6,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "This paper targets video editing by proposing a video-to-paragraph-to-video pipeline. The editing is achieved based on the modification of the paragraph. \nThe c... | {
"rating": "5;5;5;6",
"rating_avg": 5.25,
"confidence": "4;5;4;3",
"confidence_avg": 4,
"soundness": "3;2;3;3",
"soundness_avg": 2.75,
"contribution": "3;2;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": "2026-01-14T22:16:03.729703"
} | {
"id": "ZtVo8Kz4fE",
"metareview": "This paper introduces a novel video-to-paragraph-to-video pipeline for video editing, where videos are first converted into detailed textual descriptions. Users can edit these descriptions to modify video content, which is then used to regenerate the edited video. The authors fu... | {
"decision": "Reject"
} |
qpDqO7qa3R | 2407.01519v3 | DiffIR2VR-Zero: Zero-Shot Video Restoration with Diffusion-based Image Restoration Models | {
"content": "## Abstract\n\nAbstract This paper introduces a method for zero-shot video restoration using pre-trained image restoration diffusion models. Traditional video restoration methods often need retraining for different settings and struggle with limited generalization across various degradation types and da... | [
{
"id": "9EZoQ29Ns4",
"initial_rating": 5,
"confidence": 3,
"soundness": 2,
"contribution": 2,
"presentation": 2,
"summary": "This paper proposes a method for zero-shot video restoration using pre-trained image restoration diffusion models. It combines a hierarchical latent warping strat... | {
"rating": "5;5;5;6",
"rating_avg": 5.25,
"confidence": "4;3;3;5",
"confidence_avg": 3.75,
"soundness": "3;3;2;3",
"soundness_avg": 2.75,
"contribution": "3;2;2;3",
"contribution_avg": 2.5,
"presentation": "3;3;2;3",
"presentation_avg": 2.75
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:03.730475"
} | {
"id": "tgUzHA3YMA",
"metareview": "This paper presents a zero-shot video restoration method with a pre-trained image-restoration diffusion model without additional training. A hierarchical latent warping strategy is used to improve temporal consistency. The token merging technique is used with the hybrid correspo... | {
"decision": "Reject"
} |
qqZijHRcA5 | 2402.06674v3 | Impact of Dataset Properties on Membership Inference Vulnerability of Deep Transfer Learning | {
"content": "## Abstract\n\nAbstract We analyse the relationship between privacy vulnerability and dataset properties, such as examples per class and number of classes, when applying two state-of-the-art membership inference attacks (MIAs) to fine-tuned neural networks.\nWe derive per-example MIA vulnerability in te... | [
{
"id": "P4kgqrXwhU",
"initial_rating": 3,
"confidence": 4,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "This paper investigates the impact of dataset properties on membership inference attacks (MIAs). The authors analyze two state-of-the-art MIA methods, LiRA and RM... | {
"rating": "3;3;3;6",
"rating_avg": 3.75,
"confidence": "4;4;4;3",
"confidence_avg": 3.75,
"soundness": "3;3;3;4",
"soundness_avg": 3.25,
"contribution": "1;1;2;3",
"contribution_avg": 1.75,
"presentation": "2;4;3;3",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:03.731660"
} | {
"id": "RKC3gSAERM",
"metareview": "Summary: This paper studies how the number of examples per class and the number of classes impact the membership inference attacks (MIA), using two state-of-the-art MIA attacks, LiRA and RMIA. The paper finds a power-law relationship between those properties and MIA vulnerabilit... | {
"decision": "Reject"
} |
qrTOtUdz4Z | 2410.13285v1 | Composing Novel Classes: A Concept-Driven Approach to Generalized Category Discovery | {
"content": "## Abstract\n\nAbstract We tackle the generalized category discovery (GCD) problem, which aims to discover novel classes in unlabeled datasets by leveraging the knowledge of known classes. Previous works utilize the known class knowledge through shared representation spaces. Despite their progress, our ... | [
{
"id": "ToEjAgGo4Z",
"initial_rating": 5,
"confidence": 5,
"soundness": 3,
"contribution": 2,
"presentation": 2,
"summary": "This paper studies the problem of Generalized Category Discovery, which aims to assign unlabeled samples with both known and unknown class clusters. \nTo tackle t... | {
"rating": "5;5;5",
"rating_avg": 5,
"confidence": "4;4;5",
"confidence_avg": 4.333333333333333,
"soundness": "2;2;3",
"soundness_avg": 2.3333333333333335,
"contribution": "2;3;2",
"contribution_avg": 2.3333333333333335,
"presentation": "2;3;2",
"presentation_avg": 2.3333333333333335
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:03.732594"
} | {
"id": "78TqxvIUBN",
"metareview": "This paper introduces a concept learning framework for generalized category discovery (GCD) problem, which is based on the analysis experiments. The paper initially got three negative scores.\n\nThe main strengths include: 1) novel concept-based architecture; 2) well-written; 3)... | {
"decision": "Reject"
} |
qrTrnrEi9d | 2305.13582v3 | Translation and Fusion Improves Zero-shot Cross-lingual Information Extraction | {
"content": "## Abstract\n\nAbstract Large language models (LLMs) combined with instruction tuning have shown significant progress in information extraction (IE) tasks, exhibiting strong generalization capabilities to unseen datasets by following annotation guidelines.\nHowever, their applicability to low-resource l... | [
{
"id": "WkcHlyUiFr",
"initial_rating": 3,
"confidence": 4,
"soundness": 3,
"contribution": 2,
"presentation": 4,
"summary": "This work proposes a simple yet effective approach for improving cross-lingual transfer with a focus on Information Extraction (IE) tasks such as NER, relation ex... | {
"rating": "3;3;6;6",
"rating_avg": 4.5,
"confidence": "4;4;4;4",
"confidence_avg": 4,
"soundness": "3;3;3;3",
"soundness_avg": 3,
"contribution": "2;2;2;3",
"contribution_avg": 2.25,
"presentation": "2;4;3;2",
"presentation_avg": 2.75
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:03.733554"
} | {
"id": "FCeXjMAzyi",
"metareview": "This paper proposes TransFusion, a framework for improving multilingual performance in LLMs by fine-tuning models on English translations of low-resource language data and using \"annotation fusion\" (i.e., showing a fusion model data in the source language and English and annot... | {
"decision": "Reject"
} |
qssVptHTPN | 2410.11087v1 | Locality Alignment Improves Vision-Language Models | {
"content": "## Abstract\n\nAbstract Vision language models (VLMs) have seen growing adoption in recent years, but many still struggle with basic spatial reasoning errors. We hypothesize that this is due to VLMs adopting pre-trained vision backbones, specifically vision transformers (ViTs) trained with image-level s... | [
{
"id": "Bz73AUKihl",
"initial_rating": 5,
"confidence": 3,
"soundness": 2,
"contribution": 2,
"presentation": 2,
"summary": "The paper presents a method called Locality Alignment, which aims to improve the spatial reasoning capabilities of Vision Language Models (VLMs) by enhancing the ... | {
"rating": "5;5;5;6",
"rating_avg": 5.25,
"confidence": "4;4;3;4",
"confidence_avg": 3.75,
"soundness": "3;2;2;4",
"soundness_avg": 2.75,
"contribution": "2;2;2;4",
"contribution_avg": 2.5,
"presentation": "3;3;2;4",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.734414"
} | {
"id": "AeuniwpH5Y",
"metareview": "This paper proposed a locality alignment method to improve the spatial understanding ability of visual encoders so as to improve the performance of vision-language models. Based on the insight that visual encoders pre-trained on text image pairs already contains local semantics,... | {
"decision": "Accept (Poster)"
} |
qtTIP5Gjc5 | 2410.03292v1 | Demystifying the Token Dynamics of Deep Selective State Space Models | {
"content": "## Abstract\n\nAbstract Selective state space models (SSM), such as Mamba, have gained prominence for their effectiveness in modeling sequential data. Despite their outstanding empirical performance, a comprehensive theoretical understanding of deep selective SSM remains elusive, hindering their further... | [
{
"id": "I6vBIt5mfc",
"initial_rating": 5,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "This paper investigates the dynamical properties of tokens in a pre-trained Mamba model, a type of selective state space model (SSM). Despite the empirical succes... | {
"rating": "5;5;6;6",
"rating_avg": 5.5,
"confidence": "3;4;3;3",
"confidence_avg": 3.25,
"soundness": "4;2;3;3",
"soundness_avg": 3,
"contribution": "2;2;2;3",
"contribution_avg": 2.25,
"presentation": "3;2;3;3",
"presentation_avg": 2.75
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Spotlight",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.735381"
} | {
"id": "ZtpsfKjhOT",
"metareview": "This paper presents a theoretical analysis of token dynamics in deep selective ssms, particularly focusing on mamba. The key findings are: \n\n- In the one-dimensional case, tokens either converges to zero or diverge to infinity, with specific criteria determining which scenario... | {
"decision": "Accept (Spotlight)"
} |
qtWjSboqfe | 2405.15232v3 | DEEM: Diffusion models serve as the eyes of large language models for image perception | {
"content": "## Abstract\n\nAbstract The development of large language models (LLMs) has significantly advanced the emergence of large multimodal models (LMMs). While LMMs have achieved tremendous success by promoting the synergy between multimodal comprehension and creation, they often face challenges when confront... | [
{
"id": "zOaSXYyd2B",
"initial_rating": 6,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "This paper introduces the diffusion model into the image perception process of large language models to reduce excessive compression of visual details, thereby en... | {
"rating": "5;6;6;6;8",
"rating_avg": 6.2,
"confidence": "3;5;3;4;3",
"confidence_avg": 3.6,
"soundness": "2;3;3;3;3",
"soundness_avg": 2.8,
"contribution": "2;3;4;3;3",
"contribution_avg": 3,
"presentation": "2;3;3;3;3",
"presentation_avg": 2.8
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Spotlight",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.736260"
} | {
"id": "S4zyPAjATA",
"metareview": "This paper has received ratings of 8, 8, 8, 6, 6, where the reviewers have generally rated the paper positively, highlighting its innovative approach to enhancing visual robustness and mitigating hallucinations in multimodal models. \n\nThe authors has proposed a framework DEEM,... | {
"decision": "Accept (Spotlight)"
} |
qto91DryES | 2410.04213v1 | Equivariant Polynomial Functional Networks | {
"content": "## Abstract\n\nAbstract Neural Functional Networks (NFNs) have gained increasing interest due to their wide range of applications, including extracting information from implicit representations of data, editing network weights, and evaluating policies. A key design principle of NFNs is their adherence t... | [
{
"id": "L7hfbjU5Ux",
"initial_rating": 5,
"confidence": 3,
"soundness": 2,
"contribution": 2,
"presentation": 2,
"summary": "The paper proposes a new method to model neural network weights based on Equivariant Neural Functional Networks. Compared to previous approaches, this method is c... | {
"rating": "3;5;5;8",
"rating_avg": 5.25,
"confidence": "4;5;3;3",
"confidence_avg": 3.75,
"soundness": "2;2;2;4",
"soundness_avg": 2.5,
"contribution": "2;3;2;3",
"contribution_avg": 2.5,
"presentation": "2;2;2;3",
"presentation_avg": 2.25
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:03.737437"
} | {
"id": "QhxFMAJPKZ",
"metareview": "The paper is the first to propose polynomials equivariant to parameter symmetries in neural field networks, allowing for a good balance between computational complexity vs expressivity dilemma. There were also concerns regarding comparisons. There were questions regarding the pr... | {
"decision": "Reject"
} |
qtqvuBmhxU | 2410.02010v1 | MONICA: Benchmarking on Long-tailed Medical Image Classification | {
"content": "## Abstract\n\nAbstract Long-tailed learning is considered to be an extremely challenging problem in data imbalance learning. It aims to train well-generalized models from a large number of images that follow a long-tailed class distribution. In the medical field, many diagnostic imaging exams such as d... | [
{
"id": "3wj4WSpOE1",
"initial_rating": 3,
"confidence": 4,
"soundness": 2,
"contribution": 3,
"presentation": 2,
"summary": "The authors introduce the problem of long-tailed medical image classification and challenges in the field. Then they develop MONICA which is a package to benchmar... | {
"rating": "3;5;5;8",
"rating_avg": 5.25,
"confidence": "4;5;4;4",
"confidence_avg": 4.25,
"soundness": "2;3;2;4",
"soundness_avg": 2.75,
"contribution": "3;2;2;3",
"contribution_avg": 2.5,
"presentation": "2;2;2;3",
"presentation_avg": 2.25
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:03.738647"
} | {
"id": "Ih7ftv8toZ",
"metareview": "The paper introduces MONICA (Medical OpeN-source Long-taIled ClassifiCAtion), a comprehensive benchmark designed for long-tailed medical image classification (LTMIC). MONICA integrates 12 publicly available datasets across six medical domains and implements over 30 long-tailed (... | {
"decision": "Reject"
} |
qu6UMVT4k1 | 2305.01928v2 | Visual Transformation Telling | {
"content": "## Appendix A Datasheet for Visual Transformation Telling\n\n### A.1 Motivation\n\n#### A.1.1 For what purpose was the dataset created?\n\nThis dataset was created for a new visual reasoning task to test transformation reasoning ability in real-world scenarios. Given a series of states (i.e. images), VT... | [
{
"id": "Q4L8JNk6HK",
"initial_rating": 3,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 1,
"summary": "This work creates a new benchmark which tests how well the current multimodal LLMs (MLLMs) process visual transformations (i.e. change of states) between two adja... | {
"rating": "3;3;5",
"rating_avg": 3.6666666666666665,
"confidence": "4;4;3",
"confidence_avg": 3.6666666666666665,
"soundness": "2;3;2",
"soundness_avg": 2.3333333333333335,
"contribution": "1;3;2",
"contribution_avg": 2,
"presentation": "3;1;3",
"presentation_avg": 2.3333333333333335
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:03.739217"
} | {
"id": "v2y5ZIrNbW",
"metareview": "Summary: This work proposed a visual reasoning task, Visual Transformation Telling (VTT), to tell the differences between visual states transformation. A dataset with 13k samples is provided to support the task.\n\nStrengths: The experiments are comprehensive.\n\nWeaknesses: (1)... | {
"decision": "Reject"
} |
qx07JhIs8E | 2410.07675v1 | Adversarial Robustness Overestimation and Instability in TRADES | {
"content": "## Abstract\n\nAbstract This paper examines the phenomenon of probabilistic robustness overestimation in TRADES, a prominent adversarial training method. Our study reveals that TRADES sometimes yields disproportionately high PGD validation accuracy compared to the AutoAttack testing accuracy in the mult... | [
{
"id": "V9scpCWyJv",
"initial_rating": 3,
"confidence": 4,
"soundness": 2,
"contribution": 1,
"presentation": 2,
"summary": "The paper focuses on investigating TRADES which is a well known adversarial training method.\n\nThe paper identifies hyper-parameter settings in which training wi... | {
"rating": "1;3;3",
"rating_avg": 2.3333333333333335,
"confidence": "5;4;4",
"confidence_avg": 4.333333333333333,
"soundness": "1;3;2",
"soundness_avg": 2,
"contribution": "1;2;1",
"contribution_avg": 1.3333333333333333,
"presentation": "3;2;2",
"presentation_avg": 2.3333333333333335
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:03.739698"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
qxRoo7ULCo | 2406.13527v3 | 4K4DGen: Panoramic 4D Generation at 4K Resolution | {
"content": "## Abstract\n\nAbstract The blooming of virtual reality and augmented reality (VR/AR) technologies has driven an increasing demand for the creation of high-quality, immersive, and dynamic environments. However, existing generative techniques either focus solely on dynamic objects or perform outpainting ... | [
{
"id": "vA42NJmNC2",
"initial_rating": 6,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 4,
"summary": "This paper presents a method by which a static pano image is animated to produce a video pano, then subsequently lifted to 3D using 3DGS. The authors claim the f... | {
"rating": "3;6;6;8",
"rating_avg": 5.75,
"confidence": "5;4;4;3",
"confidence_avg": 4,
"soundness": "3;3;3;3",
"soundness_avg": 3,
"contribution": "3;3;3;3",
"contribution_avg": 3,
"presentation": "2;4;4;3",
"presentation_avg": 3.25
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Spotlight",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.740381"
} | {
"id": "A7N16h90Ii",
"metareview": "The paper presents an approach to lift 2D panorama images to 3D and animate them.\nIt builds on existing work for animation and depth estimation, and proposes an adoption to panoramic data.\nTo handle scarcity of training data, the paper leverages 2D diffusion models.\n\nThe rev... | {
"decision": "Accept (Spotlight)"
} |
qyU5s4fzLg | 2410.02558v1 | Improving Unsupervised Constituency Parsing via Maximizing Semantic Information | {
"content": "## Abstract\n\nAbstract Unsupervised constituency parsers organize phrases within a sentence into a tree-shaped syntactic constituent structure that reflects the organization of sentence semantics.\nHowever, the traditional objective of maximizing sentence log-likelihood (LL) does not explicitly account... | [
{
"id": "ZWq81aL9mn",
"initial_rating": 8,
"confidence": 4,
"soundness": 4,
"contribution": 4,
"presentation": 4,
"summary": "Unsupervised parsing methods typically involve maximizing sentence log likelihood, which does not correlate well with parsing accuracy. In contrast, this paper us... | {
"rating": "6;6;8;8",
"rating_avg": 7,
"confidence": "5;4;4;4",
"confidence_avg": 4.25,
"soundness": "2;3;3;4",
"soundness_avg": 3,
"contribution": "4;3;3;4",
"contribution_avg": 3.5,
"presentation": "3;3;3;4",
"presentation_avg": 3.25
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Spotlight",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.741159"
} | {
"id": "5LhYFY4pES",
"metareview": "This paper presents a novel method for learning constituency parsers in an unsupervised fashion. The reviewers are overwhelmingly supportive of acceptance as the consensus is that the paper presents a substantial improvement in this area beyond what exists in the literature. T... | {
"decision": "Accept (Spotlight)"
} |
r01fcKhzT5 | 2411.02275v1 | Breaking the Reclustering Barrier in Centroid-based Deep Clustering | {
"content": "## Abstract\n\nAbstract This work investigates an important phenomenon in centroid-based deep clustering (DC) algorithms: Performance quickly saturates after a period of rapid early gains. Practitioners commonly address early saturation with periodic reclustering, which we demonstrate to be insufficient... | [
{
"id": "llji4hNZpx",
"initial_rating": 8,
"confidence": 4,
"soundness": 3,
"contribution": 4,
"presentation": 4,
"summary": "The authors explore a (known) problem when training a clustering model from scratch as follows. The architecture has an encoder followed by a clustering head. Wit... | {
"rating": "5;6;6;8",
"rating_avg": 6.25,
"confidence": "5;4;4;4",
"confidence_avg": 4.25,
"soundness": "3;4;3;3",
"soundness_avg": 3.25,
"contribution": "2;3;2;4",
"contribution_avg": 2.75,
"presentation": "3;2;3;4",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.744354"
} | {
"id": "PEWnGLIChB",
"metareview": "This paper examines a pathological issue in certain centroid-based deep clustering methods, referred to as a reclustering barrier, and proposes a straightforward yet effective solution. While the analysis is not rigorous, the paper offers an intuitive explanation for the reclust... | {
"decision": "Accept (Poster)"
} |
r0opxuq8T8 | 2403.18142v1 | HERTA: A High-Efficiency and Rigorous Training Algorithm for Unfolded Graph Neural Networks | {
"content": "## Abstract\n\nAbstract As a variant of Graph Neural Networks (GNNs), Unfolded GNNs offer enhanced interpretability and flexibility over traditional designs.\nNevertheless, they still suffer from scalability challenges when it comes to the training cost. Although many methods have been proposed to addre... | [
{
"id": "yKpTHn0i3A",
"initial_rating": 5,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 4,
"summary": "This paper introduces HERTA, a novel training algorithm for Unfolded Graph Neural Networks (GNNs) that aims to solve scalability issues while preserving interpret... | {
"rating": "3;5;5;5",
"rating_avg": 4.5,
"confidence": "4;3;3;4",
"confidence_avg": 3.5,
"soundness": "2;3;3;3",
"soundness_avg": 2.75,
"contribution": "2;2;2;3",
"contribution_avg": 2.25,
"presentation": "2;3;1;4",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:03.748736"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
r3cWq6KKbt | 2405.16104v1 | Global Well-posedness and Convergence Analysis of Score-based Generative Models via Sharp Lipschitz Estimates | {
"content": "## Abstract\n\nAbstract We establish global well-posedness and convergence of the score-based generative models (SGM) under minimal general assumptions of initial data for score estimation. For the smooth case , we start from a Lipschitz bound of the score function with optimal time length. The optimali... | [
{
"id": "bu6SRV55OE",
"initial_rating": 5,
"confidence": 3,
"soundness": 4,
"contribution": 3,
"presentation": 3,
"summary": "This paper investigates the mathematical framework behind score-based generative models (SGMs). The authors' primary focus is to establish conditions under which ... | {
"rating": "3;5;6;6",
"rating_avg": 5,
"confidence": "2;3;3;3",
"confidence_avg": 2.75,
"soundness": "2;4;3;3",
"soundness_avg": 3,
"contribution": "2;3;3;3",
"contribution_avg": 2.75,
"presentation": "2;3;3;3",
"presentation_avg": 2.75
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.750120"
} | {
"id": "xKwjUV3WqQ",
"metareview": "The manuscript studies the score-based generative models. Compared with previous studies, the manuscript establishes sharper Lipschitz estimate of the score function, which in particular covers the non-smooth case when the data distribution might lie on lower-dimensional manifol... | {
"decision": "Accept (Poster)"
} |
r4Q86nBQka | 2410.05871v1 | A second-order-like optimizer with adaptive gradient scaling for deep learning | {
"content": "## Abstract\n\nAbstract In this empirical article, we introduce INNAprop, an optimization algorithm that combines the INNA method with the RMSprop adaptive gradient scaling. It leverages second-order information and rescaling while keeping the memory requirements of standard DL methods as AdamW or SGD w... | [
{
"id": "4AlHxJQJvN",
"initial_rating": 3,
"confidence": 3,
"soundness": 2,
"contribution": 2,
"presentation": 2,
"summary": "This paper introduces INNAprop, a combination of the INNA optimizer with RMSprop gradient scaling. The resulting method has memory requirements similar to methods... | {
"rating": "3;3;5;5",
"rating_avg": 4,
"confidence": "4;3;3;3",
"confidence_avg": 3.25,
"soundness": "4;2;2;2",
"soundness_avg": 2.5,
"contribution": "2;2;3;3",
"contribution_avg": 2.5,
"presentation": "3;2;3;3",
"presentation_avg": 2.75
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:03.751029"
} | {
"id": "XTV6BTszUk",
"metareview": "This paper studies a new optimization algorithm that combines the INNA method with the RMSprop adaptive gradient scaling. The resulting method has memory and computational requirements similar to AdamW. The paper demonstrates that this algorithm performs somewhat better than Ad... | {
"decision": "Reject"
} |
r6XqXoRT6N | 2406.16333v1 | Prompt-Consistency Image Generation (PCIG): A Unified Framework Integrating LLMs, Knowledge Graphs, and Controllable Diffusion Models | {
"content": "## Abstract\n\nAbstract The rapid advancement of Text-to-Image(T2I) generative models has enabled the synthesis of high-quality images guided by textual descriptions. Despite this significant progress, these models are often susceptible in generating contents that contradict the input text, which poses ... | [
{
"id": "mb3s0bvitI",
"initial_rating": 5,
"confidence": 5,
"soundness": 2,
"contribution": 2,
"presentation": 2,
"summary": "The paper proposes PCIG, a diffusion-based framework to significantly enhance the alignment of generated images with their\ncorresponding descriptions, addressing... | {
"rating": "3;3;5;5;5",
"rating_avg": 4.2,
"confidence": "4;5;4;4;5",
"confidence_avg": 4.4,
"soundness": "2;2;3;3;2",
"soundness_avg": 2.4,
"contribution": "1;1;2;2;2",
"contribution_avg": 1.6,
"presentation": "1;2;3;3;2",
"presentation_avg": 2.2
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:03.751729"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
r6aX67YhD9 | 2403.10553v1 | Learning to Watermark LLM-generated Text via Reinforcement Learning | {
"content": "## Abstract\n\nAbstract We study how to watermark LLM outputs, i.e. embedding algorithmically detectable signals into LLM-generated text to track misuse. Unlike the current mainstream methods that work with a fixed LLM, we expand the watermark design space by including the LLM tuning stage in the waterm... | [
{
"id": "AwJBa8Vbn9",
"initial_rating": 3,
"confidence": 3,
"soundness": 2,
"contribution": 2,
"presentation": 3,
"summary": "This paper explores embedding watermarks directly within large language models (LLMs) to detect and track misuse of generated content. Unlike prior token-level ap... | {
"rating": "3;5;5;6",
"rating_avg": 4.75,
"confidence": "3;4;4;4",
"confidence_avg": 3.75,
"soundness": "2;2;3;2",
"soundness_avg": 2.25,
"contribution": "2;2;2;2",
"contribution_avg": 2,
"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": "2026-01-14T22:16:03.752311"
} | {
"id": "tEo1MmVWlk",
"metareview": "1x reject, 2x borderline reject, and 1x borderline accept. This paper proposes a reinforcement-learning-based approach to embed watermarks into the weights of large language models, accompanied by a detector that learns to identify watermarked outputs. The reviewers agree on the... | {
"decision": "Reject"
} |
r8J3DSD5kF | 2410.17980v1 | Stick-breaking Attention | {
"content": "## Abstract\n\nAbstract The self-attention mechanism traditionally relies on the softmax operator, necessitating positional embeddings like RoPE, or position biases to account for token order.\nBut current methods using still face length generalisation challenges.\nWe propose an alternative attention me... | [
{
"id": "Ouqb5unst2",
"initial_rating": 5,
"confidence": 4,
"soundness": 4,
"contribution": 2,
"presentation": 3,
"summary": "This paper does a thorough evaluation of stick-breaking attention (also known as geometric attention) for synthetic test tasks and natural language tasks, showing... | {
"rating": "5;6;6;8",
"rating_avg": 6.25,
"confidence": "4;4;4;5",
"confidence_avg": 4.25,
"soundness": "4;3;3;3",
"soundness_avg": 3.25,
"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": "2026-01-14T22:16:03.752897"
} | {
"id": "gZzH8xfEoK",
"metareview": "The paper explores a new type of attention mechanism for synthetic test tasks and natural language tasks. The proposed attention inherently incorporates positional embeddings, eliminating the need for additional positional encoding techniques like RoPE. The method assigns a brea... | {
"decision": "Accept (Poster)"
} |
r9oqHOdoHf | 2410.10034v1 | TULIP: Token-length Upgraded CLIP | {
"content": "## Abstract\n\nAbstract We address the challenge of representing long captions in vision-language models, such as CLIP. By design these models are limited by fixed, absolute positional encodings, restricting inputs to a maximum of 77 tokens and hindering performance on tasks requiring longer description... | [
{
"id": "7xVkZBrRLF",
"initial_rating": 5,
"confidence": 4,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "The paper addresses the challenge of integrating positional information effectively in contrastive vision-language models, particularly when dealing with long cap... | {
"rating": "5;5;6",
"rating_avg": 5.333333333333333,
"confidence": "4;4;3",
"confidence_avg": 3.6666666666666665,
"soundness": "3;3;3",
"soundness_avg": 3,
"contribution": "3;2;2",
"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": "2026-01-14T22:16:03.753528"
} | {
"id": "oKRf4ibgHl",
"metareview": "This paper presents TULIP, which aims to enhance CLIP's capability to process long captions by replacing fixed positional embeddings with relative positional encodings, extending the model's inherent context window beyond the 77 token limit. After rebuttal, it received scores of... | {
"decision": "Accept (Poster)"
} |
rAylWUIKtu | 2410.09247v1 | Benchmark Inflation: Revealing LLM Performance Gaps Using Retro-Holdouts | {
"content": "## Abstract\n\nAbstract The training data for many Large Language Models (LLMs) is contaminated with test data. This means that public benchmarks used to assess LLMs are compromised, suggesting a performance gap between benchmark scores and actual capabilities. Ideally, a private holdout set could be us... | [
{
"id": "a4IqbJkkTn",
"initial_rating": 3,
"confidence": 4,
"soundness": 2,
"contribution": 2,
"presentation": 2,
"summary": "This paper introduce a criterion to create hold-out set for benchmark data to investigate data contamination issues. They introduce four rigorous tests to validat... | {
"rating": "3;3;5;6",
"rating_avg": 4.25,
"confidence": "5;4;4;4",
"confidence_avg": 4.25,
"soundness": "2;2;3;3",
"soundness_avg": 2.5,
"contribution": "2;2;3;3",
"contribution_avg": 2.5,
"presentation": "2;2;2;3",
"presentation_avg": 2.25
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:03.754181"
} | {
"id": "Y3Ahwimf94",
"metareview": "The paper introduces a retro-holdout framework to assess benchmark reliability for LLMs, which is an interesting and relevant topic in LLM evaluation. Experiments were conducted on 20 popular LLMs to demonstrate its effectiveness. However, reviewers have raised several concerns,... | {
"decision": "Reject"
} |
rBzvEEbrF7 | 2407.08296v1 | Q-GaLore: Quantized GaLore with INT4 Projection and Layer-Adaptive Low-Rank Gradients | {
"content": "## Abstract\n\nAbstract Training Large Language Models (LLMs) is memory-intensive due to the large number of parameters and associated optimization states. GaLore zhao2024galore , a recent method, reduces memory usage by projecting weight gradients into a low-rank subspace without compromising performan... | [
{
"id": "e1EubJV3r6",
"initial_rating": 6,
"confidence": 3,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "This paper improves the memory and compute efficiency of GaLore with a combination of aggressive quantization and lazy layerwise subspace exploration techniques. ... | {
"rating": "3;3;5;6",
"rating_avg": 4.25,
"confidence": "3;3;4;3",
"confidence_avg": 3.25,
"soundness": "3;2;2;3",
"soundness_avg": 2.5,
"contribution": "2;2;2;2",
"contribution_avg": 2,
"presentation": "4;3;3;3",
"presentation_avg": 3.25
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:03.754856"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
rCno6eYdXk | 2410.01611v2 | DRUPI: Dataset Reduction Using Privileged Information | {
"content": "## Abstract\n\nAbstract Dataset reduction (DR) seeks to select or distill samples from large datasets into smaller subsets while preserving performance on target tasks. Existing methods primarily focus on pruning or synthesizing data in the same format as the original dataset, typically the input data a... | [
{
"id": "J6vhJuRJyA",
"initial_rating": 3,
"confidence": 5,
"soundness": 2,
"contribution": 2,
"presentation": 2,
"summary": "This paper explores the potential to synthesize additional information beyond the standard data-label pairs as a learning target to enhance model training. The au... | {
"rating": "3;3;5;5;5",
"rating_avg": 4.2,
"confidence": "4;5;4;4;4",
"confidence_avg": 4.2,
"soundness": "2;2;3;3;2",
"soundness_avg": 2.4,
"contribution": "2;2;2;2;2",
"contribution_avg": 2,
"presentation": "3;2;3;3;3",
"presentation_avg": 2.8
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:03.755488"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
rD6LQagatR | 2407.12580v1 | E5-V: Universal Embeddings with Multimodal Large Language Models | {
"content": "## Abstract\n\nAbstract Multimodal large language models (MLLMs) have shown promising advancements in general visual and language understanding. However, the representation of multimodal information using MLLMs remains largely unexplored. In this work, we introduce a new framework, E5-V, designed to ada... | [
{
"id": "Utu5gDVo6E",
"initial_rating": 6,
"confidence": 3,
"soundness": 2,
"contribution": 3,
"presentation": 3,
"summary": "The paper introduces E5-V, a prompt-based representation designed to achieve universal multimodal embeddings based on Multimodal Large Language Models (MLLMs). E5... | {
"rating": "5;6;6;8",
"rating_avg": 6.25,
"confidence": "4;3;3;3",
"confidence_avg": 3.25,
"soundness": "3;3;2;3",
"soundness_avg": 2.75,
"contribution": "3;3;3;4",
"contribution_avg": 3.25,
"presentation": "2;3;3;4",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:03.756142"
} | {
"id": "eiGhzk2bHu",
"metareview": "This paper presents E5-V, which turns a multimodal large language models to output multimodal representations. The idea of employing single-modality training approach using text-only data achieves better multimodal embeddings is novel.\n\nThe reviewers agree the framework is int... | {
"decision": "Reject"
} |
rDRCIvTppL | 2410.10802v1 | Boosting Camera Motion Control for Video Diffusion Transformers | {
"content": "## Abstract\n\nAbstract Recent advancements in diffusion models have significantly enhanced the quality of video generation. However, fine-grained control over camera pose remains a challenge. While U-Net-based models have shown promising results for camera control, transformer-based diffusion models (D... | [
{
"id": "PLBRY6UgVA",
"initial_rating": 6,
"confidence": 5,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "This paper provides in-depth analysis on the design choices of adding camera control to U-Net-based and DiT-based video generation models. The investigated choice... | {
"rating": "5;5;5;6",
"rating_avg": 5.25,
"confidence": "5;4;3;5",
"confidence_avg": 4.25,
"soundness": "2;3;2;3",
"soundness_avg": 2.5,
"contribution": "1;2;2;3",
"contribution_avg": 2,
"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": "2026-01-14T22:16:03.756672"
} | {
"id": "vMeuNd1nnU",
"metareview": "The reviewers generally appreciated the paper’s systematic analysis of camera control in video diffusion models, highlighting the insightful exploration of conditioning methods and the introduction of Camera Motion Guidance (CMG). However, concerns were about the limited novelty... | {
"decision": "Reject"
} |
rEQqBZIz49 | 2407.05649v3 | Greener GRASS: Enhancing GNNs with Encoding, Rewiring, and Attention | {
"content": "## Abstract\n\nAbstract Graph Neural Networks (GNNs) have become important tools for machine learning on graph-structured data. In this paper, we explore the synergistic combination of graph encoding, graph rewiring, and graph attention, by introducing Gr aph A ttention with S tochastic S tructures (GRA... | [
{
"id": "k1A3T3mAHt",
"initial_rating": 6,
"confidence": 3,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "This paper proposes a random-walk-based approach to encode structural information, and rewire the graph accordingly, to enhance the performance of attention-based... | {
"rating": "3;5;5;5;6",
"rating_avg": 4.8,
"confidence": "4;4;3;4;3",
"confidence_avg": 3.6,
"soundness": "3;3;3;3;3",
"soundness_avg": 3,
"contribution": "1;3;2;2;2",
"contribution_avg": 2,
"presentation": "3;2;3;3;3",
"presentation_avg": 2.8
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.757273"
} | {
"id": "PHzCw5dLAS",
"metareview": "This paper introduces GRASS, a Graph Neural Network (GNN) architecture that integrates Relative Random Walk Probabilities (RRWP) encoding, random regular graph rewiring, and a new graph-tailored additive attention mechanism. GRASS aims to enhance long-range dependencies and sele... | {
"decision": "Accept (Poster)"
} |
rGyi8NNqB0 | 2401.14086v3 | Generating Likely Counterfactuals Using Sum-Product Networks | {
"content": "## Abstract\n\nAbstract Explainability of decisions made by AI systems is driven by both recent regulation and user demand. These decisions are often explainable only post hoc , after the fact. In counterfactual explanations, one may ask what constitutes the best counterfactual explanation. Clearly, mul... | [
{
"id": "DJbG7sJTC8",
"initial_rating": 8,
"confidence": 4,
"soundness": 4,
"contribution": 4,
"presentation": 3,
"summary": "This paper deals with the task of finding counterfactual explanations (x’) for a multi-class classification model (h(x)). Specifically, the authors aim to find a ... | {
"rating": "3;5;6;8",
"rating_avg": 5.5,
"confidence": "3;3;5;4",
"confidence_avg": 3.75,
"soundness": "3;2;3;4",
"soundness_avg": 3,
"contribution": "2;2;3;4",
"contribution_avg": 2.75,
"presentation": "2;2;3;3",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.758011"
} | {
"id": "yG4bYQ4RoX",
"metareview": "This paper is truly in the borderline where two reviewers view this positively while two are others are a bit negative about this. All the reviewers agree that the integration of SPNs and MIO to optimize multiple constraints is quite novel. The idea of learning SPNs and then fin... | {
"decision": "Accept (Poster)"
} |
rJ1xGcJVu8 | 2410.06651v1 | Toward Physics-guided Time Series Embedding | {
"content": "## Abstract\n\nAbstract In various scientific and engineering fields, the primary research areas have revolved around physics-based dynamical systems modeling and data-driven time series analysis. According to the embedding theory, dynamical systems and time series can be mutually transformed using obse... | [
{
"id": "Rn9nRVKhEY",
"initial_rating": 3,
"confidence": 2,
"soundness": 2,
"contribution": 2,
"presentation": 1,
"summary": "While the paper addresses an intriguing topic in time series modeling and proposes a novel approach, I found several aspects challenging to understand, especially... | {
"rating": "3;5;5;6",
"rating_avg": 4.75,
"confidence": "2;3;3;3",
"confidence_avg": 2.75,
"soundness": "2;3;2;4",
"soundness_avg": 2.75,
"contribution": "2;2;2;3",
"contribution_avg": 2.25,
"presentation": "1;2;1;1",
"presentation_avg": 1.25
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:03.758780"
} | {
"id": "0UOxJYifms",
"metareview": "The paper introduces a new concept \"Embedding Duality Theory\" in the context of time series analysis.\nIt aims to demonstrate that parameterized embeddings in deep learning models approximates the underlying nonlinear dynamics. \nBased on the reviews, the rebuttal and the d... | {
"decision": "Reject"
} |
rK0YJwL69S | 2408.13221v1 | Protecting against simultaneous data poisoning attacks | {
"content": "## Abstract\n\nAbstract Current backdoor defense methods are evaluated against a single attack at a time. This is unrealistic, as powerful machine learning systems are trained on large datasets scraped from the internet, which may be attacked multiple times by one or more attackers. We demonstrate that ... | [
{
"id": "9OlrMYMDEk",
"initial_rating": 6,
"confidence": 2,
"soundness": 3,
"contribution": 3,
"presentation": 2,
"summary": "The paper addresses the challenge of defending machine learning systems against multiple simultaneous data poisoning attacks, which is increasingly relevant in re... | {
"rating": "3;5;6;6;6;6",
"rating_avg": 5.333333333333333,
"confidence": "4;4;3;3;3;2",
"confidence_avg": 3.1666666666666665,
"soundness": "2;2;3;3;2;3",
"soundness_avg": 2.5,
"contribution": "2;2;2;3;3;3",
"contribution_avg": 2.5,
"presentation": "3;3;3;3;3;2",
"presentation_avg": 2.83333333333333... | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.759580"
} | {
"id": "3GAe6Y4aXc",
"metareview": "The paper presents a backdoor defense method against multiple simultaneous data poisoning attacks. Reviewers appreciated its motivation and novelty. However, reviewers raised concerns about the limited exploration of hyperparameters, the risk of hurting performance, and the effe... | {
"decision": "Accept (Poster)"
} |
rKMz6cDE7W | 2311.14652v2 | One Pass Streaming Algorithm for Super Long Token Attention Approximation in Sublinear Space | {
"content": "## Abstract\n\nAttention computation takes both the time complexity of O ( n 2 ) 𝑂 superscript 𝑛 2 O(n^{2}) and the space complexity of O ( n 2 ) 𝑂 superscript 𝑛 2 O(n^{2}) simultaneously, which makes deploying Large Language Models (LLMs) in streaming applications that involve long contexts req... | [
{
"id": "4cuSLgLoTu",
"initial_rating": 3,
"confidence": 3,
"soundness": 3,
"contribution": 1,
"presentation": 3,
"summary": "The paper proposes a new framework for doing one pass streaming algorithm instead of the quadratic attention. This is useful when the sequence length $n$ is much ... | {
"rating": "1;3;3",
"rating_avg": 2.3333333333333335,
"confidence": "5;3;3",
"confidence_avg": 3.6666666666666665,
"soundness": "2;1;3",
"soundness_avg": 2,
"contribution": "1;1;1",
"contribution_avg": 1,
"presentation": "2;1;3",
"presentation_avg": 2
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:03.760374"
} | {
"id": "jhD6nswyBQ",
"metareview": "The paper proposes a streaming algorithm for Attention computation. As mentioned by the reviewers the novelty is low and the paper cannot be accepted at this point.",
"additional_comments": "The rebuttal does not address the concerns."
} | {
"decision": "Reject"
} |
rLaMcF516k | 2410.19000v1 | Make LLMs better zero-shot reasoners: structure-oriented autonomous reasoning | {
"content": "## Abstract\n\nAbstract Zero-shot reasoning methods with Large Language Models (LLMs) offer significant advantages including great generalization to novel tasks and reduced dependency on human-crafted examples.\nHowever, the current zero-shot methods still have limitations in complex tasks, e.g., answer... | [
{
"id": "yusKrJdu6L",
"initial_rating": 3,
"confidence": 2,
"soundness": 3,
"contribution": 3,
"presentation": 4,
"summary": "The paper proposes a structure-oriented analysis method and a multi-agent reasoning system to enhance zero-shot reasoning capabilities in LLMs. The authors claim ... | {
"rating": "3;3;3;3;3;8",
"rating_avg": 3.8333333333333335,
"confidence": "4;3;4;4;2;4",
"confidence_avg": 3.5,
"soundness": "2;2;3;3;3;4",
"soundness_avg": 2.8333333333333335,
"contribution": "2;2;2;2;3;3",
"contribution_avg": 2.3333333333333335,
"presentation": "2;2;3;3;4;3",
"presentation_avg": ... | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:03.761192"
} | {
"id": "syVjmBENNK",
"metareview": "Summary: The paper introduces a method to improve zero-shot reasoning in Large Language Models (LLMs) by using a structure-oriented analysis inspired by human reasoning. This approach focuses on understanding the syntactic and grammatical structures of problem statements to enha... | {
"decision": "Reject"
} |
rPup1cWk4d | 2410.00718v1 | Pseudo-Non-Linear Data Augmentation via Energy Minimization | {
"content": "## Abstract\n\nAbstract We propose a novel and interpretable data augmentation method based on energy-based modeling and principles from information geometry . Unlike black-box generative models, which rely on deep neural networks, our approach replaces these non-interpretable transformations with expli... | [
{
"id": "IZJ0BUyB9g",
"initial_rating": 3,
"confidence": 3,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "The paper uses poset (partially-ordered set) to capture structures in the data.\nSpecifically, given a dataset, the data is modeled by a real-valued poset, where ... | {
"rating": "1;3;3;5",
"rating_avg": 3,
"confidence": "4;3;3;3",
"confidence_avg": 3.25,
"soundness": "3;3;3;2",
"soundness_avg": 2.75,
"contribution": "1;2;2;3",
"contribution_avg": 2,
"presentation": "2;2;3;2",
"presentation_avg": 2.25
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:03.762068"
} | {
"id": "er9nqcyGOA",
"metareview": "This paper explores data augmentation methods based on energy-based modelling.\n\nReviewers raised many concerns with the work, including that there was a lack of theoretical results contrary to the claim the work was theoretically grounded, that the methods section was not clea... | {
"decision": "Reject"
} |
rQ7fz9NO7f | 2410.04223v1 | Multimodal Large Language Models for Inverse Molecular Design with Retrosynthetic Planning | {
"content": "## Abstract\n\nAbstract While large language models (LLMs) have integrated images, adapting them to graphs remains challenging, limiting their applications in materials and drug design. This difficulty stems from the need for coherent autoregressive generation across texts and graphs. To address this, w... | [
{
"id": "xFutzWmDNC",
"initial_rating": 6,
"confidence": 2,
"soundness": 3,
"contribution": 3,
"presentation": 2,
"summary": "This paper propose a multimodal large language model framework to jointly model molecules and texts for drug discovery and retrosynthesis planning tasks, which is... | {
"rating": "6;6;6;6",
"rating_avg": 6,
"confidence": "4;2;5;2",
"confidence_avg": 3.25,
"soundness": "4;3;3;3",
"soundness_avg": 3.25,
"contribution": "3;2;3;3",
"contribution_avg": 2.75,
"presentation": "4;3;3;2",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.763002"
} | {
"id": "5DNaL1qhBv",
"metareview": "This paper proposes a multimodal large language model for inverse molecule design and retrosynthetic analysis. \n\nThe method is well-designed and the experiments are well-executed. The paper also proposes new datasets and benchmarks which are likely to be very useful for future... | {
"decision": "Accept (Poster)"
} |
rQ8mHhEIeB | 2406.08788v1 | Towards Understanding Link Predictor Generalizability Under Distribution Shifts | {
"content": "## Abstract\n\nAbstract Recently, multiple models proposed for link prediction (LP) demonstrate impressive results on benchmark datasets.\nHowever, many popular benchmark datasets often assume that dataset samples are drawn from the same distribution (i.e., IID samples). In real-world situations, this a... | [
{
"id": "cLedRWeJ3Q",
"initial_rating": 3,
"confidence": 4,
"soundness": 2,
"contribution": 2,
"presentation": 2,
"summary": "The paper investigates the generalizability of link prediction models under distribution shifts, a crucial problem given the structural differences often encounte... | {
"rating": "3;3;5;5;6",
"rating_avg": 4.4,
"confidence": "4;4;5;3;3",
"confidence_avg": 3.8,
"soundness": "2;2;3;2;3",
"soundness_avg": 2.4,
"contribution": "2;2;2;2;3",
"contribution_avg": 2.2,
"presentation": "1;2;4;3;3",
"presentation_avg": 2.6
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:03.763695"
} | {
"id": "P3wQ2TR068",
"metareview": "This work tackles the important task of evaluating link prediction models' generalizability under distribution shifts. The work proposes LPShift, a dataset-splitting strategy designed to induce distribution shifts based on structural properties, enabling a controlled examination... | {
"decision": "Reject"
} |
rRRgj3iIHR | 2407.10279v2 | AlphaDou: High-Performance End-to-End Doudizhu AI Integrating Bidding | {
"content": "## Abstract\n\nAbstract Artificial intelligence for card games has long been a popular topic in AI research. In recent years, complex card games like Mahjong and Texas Hold’em have been solved, with corresponding AI programs reaching the level of human experts. However, the game of Doudizhu presents sig... | [
{
"id": "jdmq61DkaJ",
"initial_rating": 3,
"confidence": 3,
"soundness": 2,
"contribution": 2,
"presentation": 2,
"summary": "This paper improves the Deep Monte Carlo for the Doudizhu Game and evaluates it by incorporating the effects of bidding(not random), which were not incorporated i... | {
"rating": "3;3;3;3",
"rating_avg": 3,
"confidence": "5;4;4;3",
"confidence_avg": 4,
"soundness": "2;2;2;2",
"soundness_avg": 2,
"contribution": "2;2;2;2",
"contribution_avg": 2,
"presentation": "2;2;2;2",
"presentation_avg": 2
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:03.764288"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
rTQNGQxm4K | 2404.04671v3 | PhyloLM: Inferring the Phylogeny of Large Language Models and Predicting their Performances in Benchmarks | {
"content": "## Abstract\n\nAbstract This paper introduces PhyloLM, a method adapting phylogenetic algorithms to Large Language Models (LLMs) to explore whether and how they relate to each other and to predict their performance characteristics. Our method calculates a phylogenetic distance metrics based on the simil... | [
{
"id": "ysd2GWVrQF",
"initial_rating": 6,
"confidence": 4,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "The paper introduces PhyloLM, a method adapting phylogenetic algorithms to Large Language Models (LLMs) to explore whether and how they relate to each other and t... | {
"rating": "3;5;6;10",
"rating_avg": 6,
"confidence": "4;3;4;4",
"confidence_avg": 3.75,
"soundness": "1;2;3;3",
"soundness_avg": 2.25,
"contribution": "2;2;2;4",
"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": "2026-01-14T22:16:03.765170"
} | {
"id": "chhW2lXgOe",
"metareview": "The paper: \n* Proposes similarity score between any two LLMs, which is, averaged over a huge set of prompts, the (normalized) dot product between string-response distributions.\n* Uses this similarity score to construct a phylogenetic tree of 100+ LLMs, showing ancestral simila... | {
"decision": "Accept (Poster)"
} |
rUvCIvI4eB | 2410.07087v2 | Towards Realistic UAV Vision-Language Navigation: Platform, Benchmark, and Methodology | {
"content": "## Abstract\n\nAbstract Developing agents capable of navigating to a target location based on language instructions and visual information, known as vision-language navigation (VLN), has attracted widespread interest.\nMost research has focused on ground-based agents, while UAV-based VLN remains relativ... | [
{
"id": "g4oxnRBYXS",
"initial_rating": 6,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "This paper makes three key contributions to advance realistic UAV vision-language navigation. First, they introduce OpenUAV, a simulation platform that properly m... | {
"rating": "6;6;6;6",
"rating_avg": 6,
"confidence": "2;3;4;3",
"confidence_avg": 3,
"soundness": "3;3;3;3",
"soundness_avg": 3,
"contribution": "3;2;3;3",
"contribution_avg": 2.75,
"presentation": "2;2;3;3",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.765973"
} | {
"id": "UevjkrmMP3",
"metareview": "The paper targets UAV-based vision-language navigation, recognizing that drones move in six degrees of freedom and therefore require continuous control, unlike ground-based robots. It presents the OpenUAV simulation platform, which offers realistic flight physics and varied envi... | {
"decision": "Accept (Poster)"
} |
rWQDzq3O5c | 2410.16699v1 | Graph Transformers Dream of Electric Flow | {
"content": "## Abstract\n\nAbstract We show theoretically and empirically that the linear Transformer, when applied to graph data, can implement algorithms that solve canonical problems such as electric flow and eigenvector decomposition. The input to the Transformer is simply the graph incidence matrix; no other e... | [
{
"id": "XnotUdpKJv",
"initial_rating": 5,
"confidence": 4,
"soundness": 2,
"contribution": 2,
"presentation": 3,
"summary": "The paper investigates the capabilities of the linear Transformer when applied to graph data, particularly its ability to implement algorithms for core graph task... | {
"rating": "5;5;6;6",
"rating_avg": 5.5,
"confidence": "4;4;2;3",
"confidence_avg": 3.25,
"soundness": "3;2;3;3",
"soundness_avg": 2.75,
"contribution": "3;2;2;2",
"contribution_avg": 2.25,
"presentation": "3;3;3;2",
"presentation_avg": 2.75
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.766777"
} | {
"id": "EVMJp1z5tU",
"metareview": "The paper explores the application of linear Transformers to graph-structured data, demonstrating both theoretical and empirical capabilities in solving canonical graph problems such as electric flow and eigenvector decomposition. The authors provide explicit weight configuratio... | {
"decision": "Accept (Poster)"
} |
rXrYdOtBfs | 2406.00894v1 | Pretrained Hybrids with MAD Skills | {
"content": "## Abstract\n\nAbstract While Transformers underpin modern large language models (LMs), there is a growing list of alternative architectures with new capabilities, promises, and tradeoffs.\nThis makes choosing the right LM architecture challenging.\nRecently-proposed hybrid architectures seek a best-of-... | [
{
"id": "clE5Ix08Xd",
"initial_rating": 6,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "The paper builds by proposing a new framework for automating the design of hybrid language models by re-using existing pre-trained models using ideas from Neural ... | {
"rating": "3;5;6",
"rating_avg": 4.666666666666667,
"confidence": "3;4;3",
"confidence_avg": 3.3333333333333335,
"soundness": "2;3;3",
"soundness_avg": 2.6666666666666665,
"contribution": "2;3;3",
"contribution_avg": 2.6666666666666665,
"presentation": "2;4;3",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:03.767707"
} | {
"id": "kKjnvHTzMC",
"metareview": "This paper introduces a novel framework for building hybrid language models from pretrained components. While reviewers praised the idea and clarity, they raised concerns on fairness of comparisons, scalability, and demonstration of true hybrid benefits over simpler baselines. T... | {
"decision": "Reject"
} |
rakhNY32vw | 2110.08057v3 | Almost Optimal Batch-Regret Tradeoff for Batch Linear Contextual Bandits | {
"content": "## Abstract\n\nAbstract We study the optimal batch-regret tradeoff for batch linear contextual bandits. For both context-blind and context-aware settings, we design batch learning algorithms and prove that they achieve the optimal regret bounds (up to logarithmic factors) for\nany batch number M 𝑀 M , ... | [
{
"id": "rDnRZVTmf2",
"initial_rating": 6,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "The paper addresses the problem of batch linear contextual bandits, a type of online learning model where decisions are made sequentially, balancing exploration a... | {
"rating": "6;6;8;8",
"rating_avg": 7,
"confidence": "4;3;3;5",
"confidence_avg": 3.75,
"soundness": "3;3;3;4",
"soundness_avg": 3.25,
"contribution": "3;3;4;4",
"contribution_avg": 3.5,
"presentation": "2;3;2;3",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.768713"
} | {
"id": "vK8ZoHGbgn",
"metareview": "A great paper that solves the remaining open problem left by Han 2020! Accept.",
"additional_comments": "NA"
} | {
"decision": "Accept (Poster)"
} |
rapXZIfwbX | 2410.12101v2 | The Persian Rug: solving toy models of superposition using large-scale symmetries | {
"content": "## Abstract\n\nAbstract We present a complete mechanistic description of the algorithm learned by a minimal non-linear sparse data autoencoder in the limit of large input dimension. The model, originally presented in Elhage et al. ( 2022 ) , compresses sparse data vectors through a linear layer and deco... | [
{
"id": "OhZatZ5BjO",
"initial_rating": 3,
"confidence": 4,
"soundness": 2,
"contribution": 2,
"presentation": 2,
"summary": "This paper characterizes the solutions of training sparse autoencoders (defined as $f(\\mathbf{x}) = \\textrm{ReLU}(W_{out} W_{in} \\mathbf{x} + \\mathbf{b})$) wi... | {
"rating": "3;3;3;5;6;6",
"rating_avg": 4.333333333333333,
"confidence": "4;3;4;3;3;3",
"confidence_avg": 3.3333333333333335,
"soundness": "2;2;2;3;4;3",
"soundness_avg": 2.6666666666666665,
"contribution": "1;2;2;3;2;3",
"contribution_avg": 2.1666666666666665,
"presentation": "2;1;2;2;4;3",
"prese... | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:03.769864"
} | {
"id": "Zla6r18Aoy",
"metareview": "The paper studies the compression of sparse data via a two-layer autoencoder with ReLU output non-linearity. Building on an ansatz for the product of the weights (supported by numerical evidence), the authors derive a closed-form expression for the population loss depending on t... | {
"decision": "Reject"
} |
rawj2PdHBq | 2410.13523v1 | Can Medical Vision-Language Pre-training Succeed with Purely Synthetic Data? | {
"content": "## Abstract\n\nAbstract Medical Vision-Language Pre-training (MedVLP) has made significant progress in enabling zero-shot tasks for medical image understanding. However, training MedVLP models typically requires large-scale datasets with paired, high-quality image-text data, which are scarce in the medi... | [
{
"id": "hUE4Mm3k42",
"initial_rating": 6,
"confidence": 4,
"soundness": 2,
"contribution": 3,
"presentation": 3,
"summary": "This paper provides a thorough evaluation of the medical visual language model with synthetic data. It is proven that training MedVLM with synthetic data and a ca... | {
"rating": "5;5;6",
"rating_avg": 5.333333333333333,
"confidence": "5;5;4",
"confidence_avg": 4.666666666666667,
"soundness": "2;3;2",
"soundness_avg": 2.3333333333333335,
"contribution": "2;3;3",
"contribution_avg": 2.6666666666666665,
"presentation": "3;1;3",
"presentation_avg": 2.333333333333333... | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:03.770896"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
rcKzU0Vns0 | 2405.11337v2 | A Unified Approach Towards Active Learning and Out-of-Distribution Detection | {
"content": "## Abstract\n\nAbstract When applying deep learning models in open-world scenarios, active learning (AL) strategies are crucial for identifying label candidates from a nearly infinite amount of unlabeled data. In this context, robust out-of-distribution (OOD) detection mechanisms are essential for handl... | [
{
"id": "SpCM8Y8233",
"initial_rating": 3,
"confidence": 3,
"soundness": 3,
"contribution": 2,
"presentation": 2,
"summary": "This paper proposes to handle active learning and out-of-distribution detection in a unified framework, SISOM. SISOM provides\na novel feature space analysis sch... | {
"rating": "1;3;3;3",
"rating_avg": 2.5,
"confidence": "5;4;5;3",
"confidence_avg": 4.25,
"soundness": "1;3;1;3",
"soundness_avg": 2,
"contribution": "1;2;1;2",
"contribution_avg": 1.5,
"presentation": "2;2;3;2",
"presentation_avg": 2.25
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:03.771623"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
reZKq6hjOZ | 2402.13901v3 | Broadening Target Distributions for Accelerated Diffusion Models via a Novel Analysis Approach | {
"content": "## Abstract\n\nAbstract The denoising diffusion model emerges recently as a powerful generative technique that converts noise into data. Theoretical convergence guarantee has been mainly studied for continuous-time diffusion models, and has been obtained for discrete-time diffusion models only for distr... | [
{
"id": "ZgIuCy9ghb",
"initial_rating": 6,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "This paper introduces a novel accelerated stochastic DDPM sampler that leverages the tilting factor representation and Tweedie's formula. This approach is designe... | {
"rating": "5;5;6",
"rating_avg": 5.333333333333333,
"confidence": "4;4;3",
"confidence_avg": 3.6666666666666665,
"soundness": "4;2;3",
"soundness_avg": 3,
"contribution": "2;2;3",
"contribution_avg": 2.3333333333333335,
"presentation": "3;1;3",
"presentation_avg": 2.3333333333333335
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.773274"
} | {
"id": "8WM86b0CpB",
"metareview": "This paper considers the problem of establishing convergence bounds for diffusion models.\nAuthors assume that score can be estimated accurately, and provide guarantees in terms of error tolerance. Authors improve the existing rates of diffusion models under finite variance cond... | {
"decision": "Accept (Poster)"
} |
rf0ZoDASnS | 2409.13931v2 | On-Device Collaborative Language Modeling via a Mixture of Generalists and Specialists | {
"content": "## Abstract\n\nAbstract On-device LLMs have gained increasing attention for their ability to enhance privacy and provide a personalized user experience.\nTo facilitate learning with private and scarce local data, federated learning has become a standard approach, though it introduces challenges related ... | [
{
"id": "cGMKpjxgVz",
"initial_rating": 6,
"confidence": 3,
"soundness": 3,
"contribution": 2,
"presentation": 4,
"summary": "This work introduces an approach for collaborative, on-device language modeling based on Federated Learning. It addresses the challenges of system and data hetero... | {
"rating": "3;3;5;6",
"rating_avg": 4.25,
"confidence": "5;4;3;3",
"confidence_avg": 3.75,
"soundness": "2;3;2;3",
"soundness_avg": 2.5,
"contribution": "1;2;2;2",
"contribution_avg": 1.75,
"presentation": "2;3;3;4",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:03.775380"
} | {
"id": "H9z8mhuKP6",
"metareview": "The paper presents a new method for building on-device language modeling. The proposed approach seeks to provide each person with a personalized language model using ideas from mixtures and federated learning. The personalized model on each device is a mixture -- built by sharin... | {
"decision": "Reject"
} |
rfrtFwnF62 | 2406.17988v1 | DICE: End-to-end Deformation Capture of Hand-Face Interactions from a Single Image | {
"content": "## Abstract\n\nAbstract Reconstructing 3D hand-face interactions with deformations from a single image is a challenging yet crucial task with broad applications in AR, VR, and gaming. The challenges stem from self-occlusions during single-view hand-face interactions, diverse spatial relationships betwee... | [
{
"id": "Q0ODZbks4m",
"initial_rating": 5,
"confidence": 4,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "This paper presents a model for the simultaneous reconstruction of face and hand meshes from a single image, based on previous work and annotated data from Shimad... | {
"rating": "5;6;6",
"rating_avg": 5.666666666666667,
"confidence": "4;5;4",
"confidence_avg": 4.333333333333333,
"soundness": "3;3;3",
"soundness_avg": 3,
"contribution": "2;2;3",
"contribution_avg": 2.3333333333333335,
"presentation": "3;3;3",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.776268"
} | {
"id": "R3V55UlJnh",
"metareview": "This work introduces an end-to-end method for capturing hand-face interaction deformations from a single image. Accurately recovering hand-face interactions with plausible deformations is significant due to its broad applications in AR/VR.\nThe proposed approach includes three k... | {
"decision": "Accept (Poster)"
} |
rgDwRdMwoS | 2410.10347v1 | A Unified Approach to Routing and Cascading for LLMs | {
"content": "## Abstract\n\nAbstract The widespread applicability of large language models (LLMs) has increased the availability of many fine-tuned models of various sizes targeting specific tasks. Given a set of such specialized models, to maximize overall performance, it is important to figure out the optimal stra... | [
{
"id": "fW9RRQUaE9",
"initial_rating": 6,
"confidence": 4,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "The paper explores the challenges of efficiently using large language models (LLMs) by evaluating existing model selection strategies: routing and cascading. Rout... | {
"rating": "3;3;5;6;8",
"rating_avg": 5,
"confidence": "3;4;4;4;3",
"confidence_avg": 3.6,
"soundness": "2;1;3;3;4",
"soundness_avg": 2.6,
"contribution": "2;1;2;2;3",
"contribution_avg": 2,
"presentation": "2;3;3;3;4",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:03.777391"
} | {
"id": "28yovUnNvH",
"metareview": "The paper studies optimal strategies for routing and cascading in LLMs. Experiments demonstrate that the proposed combined approach outperforms routing or cascading when deployed alone. \n\nReviewers appreciated the importance of the questions asked, and applauded the theoretica... | {
"decision": "Reject"
} |
rh54qNvxKO | 2403.03962v1 | Identify Critical Nodes in Complex Network with Large Language Models | {
"content": "## Abstract\n\nAbstract Identifying critical nodes in networks is a classical decision-making task, and many methods struggle to strike a balance between adaptability and utility. Therefore, we propose an approach that empowers Evolutionary Algorithm (EA) with Large Language Models (LLMs), to generate a... | [
{
"id": "vyddDysHjk",
"initial_rating": 6,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "This paper presents a framework combining Evolutionary Algorithms (EAs) and Large Language Models (LLMs) to identify critical nodes in complex networks. The EA’s ... | {
"rating": "3;3;3;5;5;6",
"rating_avg": 4.166666666666667,
"confidence": "5;3;4;3;4;3",
"confidence_avg": 3.6666666666666665,
"soundness": "2;2;3;2;3;3",
"soundness_avg": 2.5,
"contribution": "2;2;2;2;3;3",
"contribution_avg": 2.3333333333333335,
"presentation": "3;1;1;2;3;3",
"presentation_avg": 2... | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:03.778291"
} | {
"id": "n37448n0dm",
"metareview": "This paper proposed a framework combining LLMs with Evolutionary Algorithms (EA) to generate \"score_nodes\" function to identify the critical nodes in complex networks. \n\nStrengths:\n1. The idea of Integrating LLMs and EAs to address the critical node identification is inte... | {
"decision": "Reject"
} |
rhhQjGj09A | 2409.18061v1 | Optimal Protocols for Continual Learning via Statistical Physics and Control Theory | {
"content": "## Abstract\n\nAbstract Artificial neural networks often struggle with catastrophic forgetting when learning multiple tasks sequentially, as training on new tasks degrades the performance on previously learned ones. Recent theoretical work has addressed this issue by analysing learning curves in synthet... | [
{
"id": "SauxVReo8B",
"initial_rating": 5,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "The paper proposes an ODE relating network parameters, training control parameters (such as learning rate and which task to learn on), and the final performance o... | {
"rating": "5;5;8;8",
"rating_avg": 6.5,
"confidence": "4;3;4;4",
"confidence_avg": 3.75,
"soundness": "3;2;4;3",
"soundness_avg": 3,
"contribution": "2;3;3;4",
"contribution_avg": 3,
"presentation": "3;2;4;4",
"presentation_avg": 3.25
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.779301"
} | {
"id": "NMaUXX9bxC",
"metareview": "This paper theoretically approaches continual learning using ODEs (for learning curves of online SGD), applying it to teacher-student continual learning methods, which is novel. They derive an optimal learning rate schedule and study the influence of task similarity on forgettin... | {
"decision": "Accept (Poster)"
} |
rj7wUcLgfw | 2405.14522v1 | Explaining Black-box Model Predictions via Two-level Nested Feature Attributions with Consistency Property | {
"content": "## Abstract\n\nAbstract Techniques that explain the predictions of black-box machine learning models are crucial to make the models transparent, thereby increasing trust in AI systems.\nThe input features to the models often have a nested structure that consists of high- and low-level features, and each... | [
{
"id": "QQh2trspDu",
"initial_rating": 5,
"confidence": 3,
"soundness": 2,
"contribution": 3,
"presentation": 3,
"summary": "The authors propose a method to get high level and low level feature attributions for image classification in multiple instance learning and text classification w... | {
"rating": "3;3;5",
"rating_avg": 3.6666666666666665,
"confidence": "3;4;3",
"confidence_avg": 3.3333333333333335,
"soundness": "2;1;2",
"soundness_avg": 1.6666666666666667,
"contribution": "2;1;3",
"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": "2026-01-14T22:16:03.780179"
} | {
"id": "y0zuqT4iLB",
"metareview": "This paper introduces a model-agnostic explanation method for simultaneously estimating high-level and low-level feature attributions (HiFAs and LoFAs) in image and text classification tasks. The method uses a LIME-like local surrogate model with a consistency constraint to ensu... | {
"decision": "Reject"
} |
rjuZyMfLSd | 2407.00717v1 | Learning system dynamics without forgetting | {
"content": "## Abstract\n\nAbstract Predicting the trajectories of systems with unknown dynamics ( i.e. the governing rules) is crucial in various research fields, including physics and biology. This challenge has gathered significant attention from diverse communities. Most existing works focus on learning fixed s... | [
{
"id": "Yq6Cr9Clo2",
"initial_rating": 5,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "The paper proposed a continual learning framework based on GraphODE over dynamical systems. The key motivation is to address the catastrophic forgetting problem i... | {
"rating": "5;5;5;6",
"rating_avg": 5.25,
"confidence": "3;4;4;4",
"confidence_avg": 3.75,
"soundness": "3;2;3;2",
"soundness_avg": 2.5,
"contribution": "2;2;3;2",
"contribution_avg": 2.25,
"presentation": "2;3;3;3",
"presentation_avg": 2.75
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.781092"
} | {
"id": "kFl0SBIc8v",
"metareview": "In this work, authors introduce Mode-switching Graph ODE (MS-GODE), a novel approach for Continual Dynamics Learning (CDL) that aims to model systems with evolving dynamics while preventing catastrophic forgetting. The work systematically investigates the CDL problem, proposes a... | {
"decision": "Accept (Poster)"
} |
rkzabmWl5k | 2410.07718v2 | Hallo2: Long-Duration and High-Resolution Audio-Driven Portrait Image Animation | {
"content": "## Abstract\n\nAbstract Recent advances in latent diffusion-based generative models for portrait image animation, such as Hallo, have achieved impressive results in short-duration video synthesis.\nIn this paper, we present updates to Hallo, introducing several design enhancements to extend its capabili... | [
{
"id": "oLy6e6CThJ",
"initial_rating": 6,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "This paper presents a pipeline for generating long-duration, high-resolution, audio-driven facial animations. To enable extended video generation, the authors int... | {
"rating": "5;5;6;6",
"rating_avg": 5.5,
"confidence": "4;4;5;4",
"confidence_avg": 4.25,
"soundness": "3;3;3;3",
"soundness_avg": 3,
"contribution": "2;2;1;3",
"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": "2026-01-14T22:16:03.782001"
} | {
"id": "gyPD41kvxP",
"metareview": "The paper addresses the problem of audio-driven portrait image animation and proposes a series of improvements which enable long duration video, high resolution (4k), and incorporate adjustable semantic textual labels for portrait expressions. The long duration video is enabled ... | {
"decision": "Accept (Poster)"
} |
rlgplAuN2p | 2410.23703v1 | OCEAN: Offline Chain-of-thought Evaluation and Alignment in Large Language Models | {
"content": "## Abstract\n\nAbstract Offline evaluation of LLMs is crucial in understanding their capacities, though current methods remain underexplored in existing research.\nIn this work, we focus on the offline evaluation of the chain-of-thought capabilities and show how to optimize LLMs based on the proposed ev... | [
{
"id": "9F5qiEIRSu",
"initial_rating": 6,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "The paper explores a novel approach for enhancing the chain-of-thought (CoT) reasoning in large language models (LLMs) through an offline evaluation and optimizat... | {
"rating": "3;6;6;6;6",
"rating_avg": 5.4,
"confidence": "3;4;4;3;3",
"confidence_avg": 3.4,
"soundness": "3;3;3;3;3",
"soundness_avg": 3,
"contribution": "2;3;3;3;3",
"contribution_avg": 2.8,
"presentation": "2;3;3;4;3",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.782785"
} | {
"id": "fl35raoCDo",
"metareview": "This paper proposes a way to do an offline evaluation of the chain-of-thought capabilities of LLM using a knowledge graph and an inverse-propensity-based metric. Reviewers are generally positive on this one and found the idea of aligning knowledge graph reasoning and chain-of-th... | {
"decision": "Accept (Poster)"
} |
rn8r7GqJm6 | 2410.05530v1 | VisDiff: SDF-Guided Polygon Generation for Visibility Reconstruction and Recognition | {
"content": "## Abstract\n\nAbstract The capability to learn latent representations plays a key role in the effectiveness of recent machine learning methods.\nAn active frontier in representation learning is understanding representations for combinatorial structures which may not admit well-behaved local neighborhoo... | [
{
"id": "vLnwdnRbUV",
"initial_rating": 5,
"confidence": 2,
"soundness": 3,
"contribution": 2,
"presentation": 2,
"summary": "The paper presents a diffusion model for 2D polygonal shapes represented in terms of their visibility graph (a skeleton-like structure obtained by taking the dual... | {
"rating": "3;5;5;5;6;8",
"rating_avg": 5.333333333333333,
"confidence": "3;4;3;2;3;2",
"confidence_avg": 2.8333333333333335,
"soundness": "1;2;3;3;2;3",
"soundness_avg": 2.3333333333333335,
"contribution": "2;2;2;2;3;3",
"contribution_avg": 2.3333333333333335,
"presentation": "2;2;2;2;2;4",
"prese... | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:03.783630"
} | {
"id": "BS6TNEnkYw",
"metareview": "The paper studies the problem of reconstructing a polygon from its visibility graph. The main contribution of the paper is a diffusion based approach for the problem, including the development of a loss function and curated dataset for training.\n\nThe reviewers found the applic... | {
"decision": "Reject"
} |
rpEATZvmjr | 2311.12548v1 | Multi-Session Budget Optimization for Forward Auction-based Federated Learning | {
"content": "## Abstract\n\nAbstract Auction-based Federated Learning (AFL) has emerged as an important research field in recent years. The prevailing strategies for FL model users (MUs) assume that the entire team of the required data owners (DOs) for an FL task must be assembled before training can commence. In pr... | [
{
"id": "rS8fT0fOys",
"initial_rating": 6,
"confidence": 2,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "Auction-based federated learning consists of three types of participant: 1) data owners, who are willing to share their potentially sensitive data if they are app... | {
"rating": "3;3;5;6",
"rating_avg": 4.25,
"confidence": "3;4;2;2",
"confidence_avg": 2.75,
"soundness": "2;2;2;3",
"soundness_avg": 2.25,
"contribution": "2;2;2;3",
"contribution_avg": 2.25,
"presentation": "1;3;3;3",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:03.784643"
} | {
"id": "9EHjqhrWsV",
"metareview": "The paper studies an auction-based FL setting, particularly one in which auctions span multiple sequential sessions. An RL algorithm is proposed to jointly optimize budget pacing and bidding across sessions, and evaluated experimentally.\n\nStrengths included the thorough experi... | {
"decision": "Reject"
} |
rpbzBXdo4x | 2410.21333v3 | Mind Your Step (by Step): Chain-of-Thought can Reduce Performance on Tasks where Thinking Makes Humans Worse | {
"content": "## Abstract\n\nAbstract Chain-of-thought (CoT) prompting has become a widely used strategy for working with large language and multimodal models. While CoT has been shown to improve performance across many tasks,\ndetermining the settings in which it is effective remains an ongoing effort. In particular... | [
{
"id": "Rlgzt1uBoZ",
"initial_rating": 5,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 1,
"summary": "Inspired by studies of tasks where explicit thinking hurts human performance, this paper studies when chain of thought (CoT) can hurt performance in LLMs and VLMs... | {
"rating": "3;5;5;6",
"rating_avg": 4.75,
"confidence": "4;4;3;4",
"confidence_avg": 3.75,
"soundness": "2;2;3;3",
"soundness_avg": 2.5,
"contribution": "3;2;3;3",
"contribution_avg": 2.75,
"presentation": "4;2;1;3",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:03.785397"
} | {
"id": "KeDIjp6Y2r",
"metareview": "This paper investigates when CoT prompting reduces LLM performance by drawing parallels to cases where verbal thinking impairs human performance. While the core idea is interesting and results show some significant performance drops with CoT, reviewers raised valid concerns abou... | {
"decision": "Reject"
} |
rsZwwjYHuD | 2408.02032v2 | Self-Introspective Decoding: Alleviating Hallucinations for Large Vision-Language Models | {
"content": "## Abstract\n\nAbstract Hallucination remains a significant challenge in Large Vision-Language Models (LVLMs).\nTo alleviate this issue, some methods, known as contrastive decoding, induce hallucinations by manually disturbing the raw vision or instruction inputs and then mitigate them by contrasting th... | [
{
"id": "KoYWtYbTyK",
"initial_rating": 8,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "This manuscript proposes to solve the hallucination issue in Large Vision-Language Models (LVLMs). The proposed method named Self-Introspective Decoding (SID) aim... | {
"rating": "3;5;6;8",
"rating_avg": 5.5,
"confidence": "4;5;4;3",
"confidence_avg": 4,
"soundness": "2;3;3;3",
"soundness_avg": 2.75,
"contribution": "2;3;3;3",
"contribution_avg": 2.75,
"presentation": "3;3;3;3",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.786239"
} | {
"id": "S4TmvfdEn1",
"metareview": "This work brings a new method, named Self-Introspective Decoding (SID), that alleviates hallucinations in large vision language models. Existing approaches either introduce noise or need double inference costs, while this paper provides an alternative that uses a Context and Tex... | {
"decision": "Accept (Poster)"
} |
rtUjj03qZv | 2410.05767v1 | Grounding is All You Need? Dual Temporal Grounding for Video Dialog | {
"content": "## Abstract\n\nAbstract In the realm of video dialog response generation, the understanding of video content and the temporal nuances of conversation history are paramount. While a segment of current research leans heavily on large-scale pretrained visual-language models and often overlooks temporal dyn... | [
{
"id": "u5yPl7Rilr",
"initial_rating": 5,
"confidence": 3,
"soundness": 3,
"contribution": 2,
"presentation": 2,
"summary": "Previous approaches for video dialog have rarely grounded relevant video frames, and more specifically, relevant dialog turns for generating a response to the cur... | {
"rating": "3;3;5;5",
"rating_avg": 4,
"confidence": "5;4;3;3",
"confidence_avg": 3.75,
"soundness": "2;3;2;3",
"soundness_avg": 2.5,
"contribution": "2;3;2;2",
"contribution_avg": 2.25,
"presentation": "2;2;1;2",
"presentation_avg": 1.75
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:03.786925"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
rvhu4V7yrX | 2306.04169v2 | Efficient Alternating Minimization with Applications to Weighted Low Rank Approximation | {
"content": "## Abstract\n\nWeighted low rank approximation is a fundamental problem in numerical linear algebra, and it has many applications in machine learning. Given a matrix M ∈ ℝ n × n 𝑀 superscript ℝ 𝑛 𝑛 M\\in\\mathbb{R}^{n\\times n} , a weight matrix W ∈ ℝ ≥ 0 n × n 𝑊 superscript subscript ℝ absent 0 𝑛 ... | [
{
"id": "K6XFOXQZg7",
"initial_rating": 6,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "This paper studies the problem of weighted low-rank matrix approximation. In this problem one is given a weight matrix W\n and observes a matrix M', and the goal ... | {
"rating": "5;6;6;6",
"rating_avg": 5.75,
"confidence": "4;4;2;4",
"confidence_avg": 3.5,
"soundness": "3;3;3;3",
"soundness_avg": 3,
"contribution": "3;3;3;3",
"contribution_avg": 3,
"presentation": "2;3;3;3",
"presentation_avg": 2.75
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.787950"
} | {
"id": "Le6KBU3MNS",
"metareview": "This paper introduces a new framework for weighted low-rank approximation based on approximate alternating updates. With the proposed approach, the authors show that the algorithm’s runtime is reduced from $\\|W\\|_0 k^2$ to $\\|W\\|_0 k$ , where \n$k$ is the rank of the sought ... | {
"decision": "Accept (Poster)"
} |
rwAEQWEqkX | 2410.02746v1 | Contrastive Localized Language-Image Pre-Training | {
"content": "## Abstract\n\nAbstract Contrastive Language-Image Pre-training (CLIP) has been a celebrated method for training vision encoders to generate image/text representations facilitating various applications. Recently, CLIP has been widely adopted as the vision backbone of multimodal large language models (ML... | [
{
"id": "LdcGyNJEjC",
"initial_rating": 6,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "This paper addresses a critical limitation of CLIP by introducing CLOC (Contrastive Localized Language-Image Pre-training), which enhances CLIP's capability to un... | {
"rating": "1;6;6;6",
"rating_avg": 4.75,
"confidence": "5;3;3;4",
"confidence_avg": 3.75,
"soundness": "3;3;2;3",
"soundness_avg": 2.75,
"contribution": "1;3;2;3",
"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": "2026-01-14T22:16:03.789263"
} | {
"id": "BvqQcwIbDt",
"metareview": "This paper proposes a pretraining method to enhance the localization ability of contrastive image-text pretraining (i.e., CLIP), which is considered an important topic in the research field as CLIP models are widely used across various applications. The paper presents a new meth... | {
"decision": "Reject"
} |
rwNzSB3sDt | 2402.09240v2 | Switch EMA: A Free Lunch for Better Flatness and Sharpness | {
"content": "## Abstract\n\nAbstract Exponential Moving Average (EMA) is a widely used weight averaging (WA) regularization to learn flat optima for better generalizations without extra cost in deep neural network (DNN) optimization.\nDespite achieving better flatness, existing WA methods might fall into worse final... | [
{
"id": "NVThUp4kqT",
"initial_rating": 3,
"confidence": 2,
"soundness": 2,
"contribution": 2,
"presentation": 1,
"summary": "The paper proposes SEMA methods, which apply exponential moving average (EMA) periodically at specific intervals to improve model \"sharpness\" and \"flatness.\" ... | {
"rating": "3;3;6;8",
"rating_avg": 5,
"confidence": "4;2;4;3",
"confidence_avg": 3.25,
"soundness": "3;2;4;4",
"soundness_avg": 3.25,
"contribution": "2;2;4;4",
"contribution_avg": 3,
"presentation": "3;1;3;3",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:03.790225"
} | {
"id": "8Cuy9L6zjq",
"metareview": "The paper presents the switch exponential moving average (SEMA) to better train deep-net models. The high-level idea is to switch between using the original model parameter vs. the SEMA parameter at a selected interval. The paper demonstrates experiments on a large set of domain... | {
"decision": "Reject"
} |
rxVvRBgqmS | 2406.09326v1 | PianoMotion10M: Dataset and Benchmark for Hand Motion Generation in Piano Performance | {
"content": "## Abstract\n\nAbstract Recently, artificial intelligence techniques for education have been received increasing attentions, while it still remains an open problem to design the effective music instrument instructing systems. Although key presses can be directly derived from sheet music, the transitiona... | [
{
"id": "VOK8wCFx3a",
"initial_rating": 6,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "This paper presents PianoMotion10M, a large-scale dataset designed to advance hand motion generation in piano performance, featuring 116 hours of piano videos ann... | {
"rating": "6;6;6;6",
"rating_avg": 6,
"confidence": "4;5;3;4",
"confidence_avg": 4,
"soundness": "3;3;2;3",
"soundness_avg": 2.75,
"contribution": "3;3;3;3",
"contribution_avg": 3,
"presentation": "3;3;3;3",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Spotlight",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.791480"
} | {
"id": "8FOAnAVB4U",
"metareview": "This paper introduces PianoMotion10M, a new dataset of piano performances with video, audio, MIDI, and hand motion data. The dataset contains over 116 hours of piano videos annotated with 10 million hand poses. The authors also introduce a baseline model that generates realisti... | {
"decision": "Accept (Spotlight)"
} |
ryIHtXE9uG | 2410.24087v1 | In-context Fine-tuning for Time-series Foundation Models | {
"content": "## Abstract\n\nAbstract 0 0 footnotetext: Authors listed in alphabetical order. Motivated by the recent success of time-series foundation models for zero-shot forecasting, we present a methodology for in-context fine-tuning of a time-series foundation model. In particular, we design a pretrained foundat... | [
{
"id": "6meFYckLT3",
"initial_rating": 6,
"confidence": 4,
"soundness": 4,
"contribution": 3,
"presentation": 1,
"summary": "The paper proposes the concept of in-context fine-tuning for time series forecasting. The approach proposes to leverage an existing autoregressive transformer bas... | {
"rating": "3;5;6;6;8",
"rating_avg": 5.6,
"confidence": "5;4;4;4;3",
"confidence_avg": 4,
"soundness": "1;3;3;4;3",
"soundness_avg": 2.8,
"contribution": "2;2;3;3;4",
"contribution_avg": 2.8,
"presentation": "2;3;3;1;3",
"presentation_avg": 2.4
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:03.792063"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
ryKrRCbcCX | 2410.20199v1 | Rethinking the Uncertainty: A Critical Review and Analysis in the Era of Large Language Models | {
"content": "## Abstract\n\nAbstract In recent years, Large Language Models (LLMs) have become fundamental to a broad spectrum of artificial intelligence applications. As the use of LLMs expands, precisely estimating the uncertainty in their predictions has become crucial. Current methods often struggle to accuratel... | [
{
"id": "qjsJK3Q94d",
"initial_rating": 3,
"confidence": 3,
"soundness": 2,
"contribution": 2,
"presentation": 3,
"summary": "This paper looks uncertainties specific to LLM and designs a comprehensive framework to identify the sources and types of them. The authors demonstrate how uncert... | {
"rating": "3;3;3;8",
"rating_avg": 4.25,
"confidence": "4;2;3;4",
"confidence_avg": 3.25,
"soundness": "2;2;2;4",
"soundness_avg": 2.5,
"contribution": "2;2;2;3",
"contribution_avg": 2.25,
"presentation": "2;2;3;4",
"presentation_avg": 2.75
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:03.792802"
} | {
"id": "NPQ7MkOJYZ",
"metareview": "This paper proposed a new framework for uncertainty in LLMs. To this end, it considered standardized definitions and a taxonomy for various types of uncertainty that were previously considered in the literature. The paper then presents a comparison of different existing methods ... | {
"decision": "Reject"
} |
rzx3vcvlzj | 2410.01469v1 | TIGER: Time-frequency Interleaved Gain Extraction and Reconstruction for Efficient Speech Separation | {
"content": "## Abstract\n\nAbstract In recent years, much speech separation research has focused primarily on improving model performance. However, for low-latency speech processing systems, high efficiency is equally important. Therefore, we propose a speech separation model with significantly reduced parameters a... | [
{
"id": "J2qQxn7cWJ",
"initial_rating": 6,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "The paper developed a deep learning model for speech separation; and it not only focuses on improving model performance but model efficiency. The model is lightwe... | {
"rating": "1;5;6;6;8;8",
"rating_avg": 5.666666666666667,
"confidence": "5;3;4;3;3;4",
"confidence_avg": 3.6666666666666665,
"soundness": "1;3;3;3;4;4",
"soundness_avg": 3,
"contribution": "1;2;3;3;4;4",
"contribution_avg": 2.8333333333333335,
"presentation": "3;3;3;3;4;3",
"presentation_avg": 3.1... | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.794136"
} | {
"id": "XCSsEw078W",
"metareview": "The authors presented a deep learning model for speech separation that emphasizes not only improving performance but also enhancing model efficiency. The proposed model is lightweight and incorporates a novel band-split strategy along with a new frequency-frame interleaved (FFI)... | {
"decision": "Accept (Poster)"
} |
s0gdfKcmoU | 2406.04201v2 | Securing Equal Share: A Principled Approach for Learning Multiplayer Symmetric Games | {
"content": "## Abstract\n\nAbstract This paper examines multiplayer symmetric constant-sum games with more than two players in a competitive setting, including examples like Mahjong, Poker, and various board and video games. In contrast to two-player zero-sum games, equilibria in multiplayer games are neither uniqu... | [
{
"id": "gihlKavVVX",
"initial_rating": 6,
"confidence": 3,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "The paper considers multiplayer symmetric constant-sum games, focusing on achieving equal share - where a player secures an expected payoff of $C/n$ in an n-playe... | {
"rating": "5;5;6;6",
"rating_avg": 5.5,
"confidence": "4;4;3;3",
"confidence_avg": 3.5,
"soundness": "3;3;4;3",
"soundness_avg": 3.25,
"contribution": "2;2;3;2",
"contribution_avg": 2.25,
"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": "2026-01-14T22:16:03.795284"
} | {
"id": "2bAOMX5aZx",
"metareview": "The paper proposes a framework for multiplayer symmetric constant-sum games by introducing the concept of equal share, where a player secures an expected payoff of C/n in an n-player game with total payoff C. The reviewers generally found the problem well-motivated and welcomed ... | {
"decision": "Reject"
} |
s15HrqCqbr | 2410.09156v1 | On Discriminative Probabilistic Modeling for Self-Supervised Representation Learning | {
"content": "## Abstract\n\nAbstract We study the discriminative probabilistic modeling problem on a continuous domain for (multimodal) self-supervised representation learning. To address the challenge of computing the integral in the partition function for each anchor data, we leverage the multiple importance sampl... | [
{
"id": "fDS2pMKkGD",
"initial_rating": 5,
"confidence": 4,
"soundness": 2,
"contribution": 3,
"presentation": 3,
"summary": "This paper explores discriminative probabilistic modeling in unimodal and multimodal self-supervised learning. By applying multiple importance sampling to Monte C... | {
"rating": "5;6;8",
"rating_avg": 6.333333333333333,
"confidence": "4;4;3",
"confidence_avg": 3.6666666666666665,
"soundness": "2;3;3",
"soundness_avg": 2.6666666666666665,
"contribution": "3;3;3",
"contribution_avg": 3,
"presentation": "3;3;3",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.796466"
} | {
"id": "VLr1ayswnD",
"metareview": "The paper introduces a non-parametric method leveraging multiple importance sampling (MIS) to enhance self-supervised representation learning, tackling the limitations associated with the InfoNCE loss. By utilizing multiple importance sampling for Monte Carlo integration, the ap... | {
"decision": "Accept (Poster)"
} |
s1kyHkdTmi | 2410.13166v2 | An Evolved Universal Transformer Memory | {
"content": "## Abstract\n\nAbstract Prior methods propose to offset the escalating costs of modern foundation models by dropping specific parts of their contexts with hand-designed rules, while attempting to preserve their original performance.\nWe overcome this trade-off with Neural Attention Memory Models (NAMMs)... | [
{
"id": "6oZtgvIgz2",
"initial_rating": 5,
"confidence": 4,
"soundness": 3,
"contribution": 2,
"presentation": 2,
"summary": "They propose a learned strategy for maintaining the size of your KV cache. An auxiliary model utilizes the attention map to produce a score of how important each ... | {
"rating": "3;5;6;6",
"rating_avg": 5,
"confidence": "3;4;4;4",
"confidence_avg": 3.75,
"soundness": "2;3;2;3",
"soundness_avg": 2.5,
"contribution": "2;2;3;3",
"contribution_avg": 2.5,
"presentation": "2;2;3;3",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.797629"
} | {
"id": "79rayUFHcr",
"metareview": "Post rebuttal, all of reviewers vote for acceptance. The AC checked all the materials and concurs that the paper has done a valuable exploration of using evolution algorithms to compress the KV cache based Transformer memory for efficient decoding. The paper received initial mix... | {
"decision": "Accept (Poster)"
} |
s1zO0YBEF8 | 2410.08309v1 | Dynamics of Concept Learning and Compositional Generalization | {
"content": "## Abstract\n\nAbstract Prior work has shown that text-conditioned diffusion models can learn to identify and manipulate primitive concepts underlying a compositional data-generating process, enabling generalization to entirely novel, out-of-distribution compositions.\nBeyond performance evaluations, th... | [
{
"id": "wKdyl5BbLC",
"initial_rating": 8,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 2,
"summary": "The paper sets out to explain previously reported characteristics of the learning dynamics of text-conditioned diffusion models with respect to their ability to ... | {
"rating": "3;6;6;6",
"rating_avg": 5.25,
"confidence": "4;4;3;3",
"confidence_avg": 3.5,
"soundness": "2;3;3;3",
"soundness_avg": 2.75,
"contribution": "1;3;3;3",
"contribution_avg": 2.5,
"presentation": "2;3;3;3",
"presentation_avg": 2.75
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.798605"
} | {
"id": "nDcleV8C4L",
"metareview": "(a) summary\n\nThis paper studies how to model the learning dynamics of compositional generalization(CG) in Neural Networks. It introduces a structured identity mapping (SIM) task for analyzing the learning dynamics mathematically. SIM helps to explain the empirical observations... | {
"decision": "Accept (Poster)"
} |
s20W12XTF8 | 2410.02298v2 | Jailbreak Antidote: Runtime Safety-Utility Balance via Sparse Representation Adjustment in Large Language Models | {
"content": "## Abstract\n\nAbstract As large language models (LLMs) become integral to various applications, ensuring both their safety and utility is paramount. Jailbreak attacks, which manipulate LLMs into generating harmful content, pose significant challenges to this balance. Existing defenses, such as prompt e... | [
{
"id": "MyBvaNxXM4",
"initial_rating": 5,
"confidence": 3,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "This paper proposed a new defense to mitigate jailbreak attacks. The key idea of the proposed defense is to modify the internal state of an LLM during inference. ... | {
"rating": "5;5;5;8",
"rating_avg": 5.75,
"confidence": "4;4;3;3",
"confidence_avg": 3.5,
"soundness": "2;2;3;3",
"soundness_avg": 2.5,
"contribution": "2;2;2;3",
"contribution_avg": 2.25,
"presentation": "3;3;3;4",
"presentation_avg": 3.25
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.799478"
} | {
"id": "Ht5WIoZe7o",
"metareview": "The submission received the ratings of four reviewers, which recommended 5, 6, 8 and 6, averaging 6.25. Given the plenty of competitive submissions in ICLR, this stands at a score above the borderline. The reviewers' concerns focus on the insufficient baselines, unclear technica... | {
"decision": "Accept (Poster)"
} |
s3003xWtfd | 2410.18311v1 | CoreInfer: Accelerating Large Language Model Inference with Semantics-Inspired Adaptive Sparse Activation | {
"content": "## Abstract\n\nAbstract Large language models (LLMs) with billions of parameters have sparked a new wave of exciting AI applications. However, their high computational costs and memory demands during inference pose significant challenges. Adaptive sparse activation inference, which activates only a smal... | [
{
"id": "HvZzezH6fW",
"initial_rating": 3,
"confidence": 4,
"soundness": 2,
"contribution": 2,
"presentation": 2,
"summary": "This work proposes a new sparse activation inference method based on the semantics of input sentences. It finds that the activated neurons remain relatively stabl... | {
"rating": "3;6;6;8",
"rating_avg": 5.75,
"confidence": "4;3;5;4",
"confidence_avg": 4,
"soundness": "2;2;3;4",
"soundness_avg": 2.75,
"contribution": "2;2;3;4",
"contribution_avg": 2.75,
"presentation": "2;3;2;4",
"presentation_avg": 2.75
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:03.800286"
} | {
"id": "FEdWC3Urg6",
"metareview": "**Summary:** The paper introduces CoreInfer, a method for accelerating LLM inference through semantics-inspired adaptive sparse activation. CoreInfer predicts core neurons critical for sentence semantics during the pre-filling stage and removes other neurons during subsequent in... | {
"decision": "Reject"
} |
s3FTX4Ay55 | 2410.13882v1 | Articulate-Anything: Automatic Modeling of Articulated Objects via a Vision-Language Foundation Model | {
"content": "## Abstract\n\nAbstract Interactive 3D simulated objects are crucial in AR/VR, animations, and robotics, driving immersive experiences and advanced automation. However, creating these articulated objects requires extensive human effort and expertise, limiting their broader applications. To overcome this... | [
{
"id": "YgofRilvka",
"initial_rating": 8,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "This paper presents a method for generating simulated 3D objects, with moving parts, that can be inferred with from multi-modal inputs (text, images, videos). Thi... | {
"rating": "3;5;6;8;8",
"rating_avg": 6,
"confidence": "4;3;3;5;3",
"confidence_avg": 3.6,
"soundness": "2;3;2;4;3",
"soundness_avg": 2.8,
"contribution": "2;2;2;4;3",
"contribution_avg": 2.6,
"presentation": "3;3;3;3;3",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.800889"
} | {
"id": "H7zCIs8PZT",
"metareview": "This work introduces a novel approach to automate the articulation of complex objects from multi-modal inputs (text, images, videos). Four reviewers have expressed positive feedback, while one reviewer has raised some concerns. The author has done a good job during rebuttal. The... | {
"decision": "Accept (Poster)"
} |
s3IBHTTDYl | 2405.20131v2 | Language Models Need Inductive Biases to Count Inductively | {
"content": "## Abstract\n\nAbstract Counting is a fundamental example of generalization, whether viewed through the mathematical lens of Peano’s axioms defining the natural numbers or the cognitive science literature for children learning to count. The argument holds for both cases that learning to count means lear... | [
{
"id": "uVbdg7K77f",
"initial_rating": 8,
"confidence": 4,
"soundness": 4,
"contribution": 4,
"presentation": 4,
"summary": "This paper studies the ability of language models to learn counting inductively. Specifically, whether they can generalize counting beyond their training data. Th... | {
"rating": "3;5;8;8",
"rating_avg": 6,
"confidence": "5;4;4;4",
"confidence_avg": 4.25,
"soundness": "2;3;4;4",
"soundness_avg": 3.25,
"contribution": "1;2;4;4",
"contribution_avg": 2.75,
"presentation": "2;2;4;4",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.801613"
} | {
"id": "BfFORU8pHW",
"metareview": "This paper investigates the claim that Transformers can count. Through careful experimentation, it shows that they can’t, like modern RNN architectures, while traditional RNN architectures can. This paper shows a surprising result about a fundamental capacity of sequence models ... | {
"decision": "Accept (Poster)"
} |
s3sJenvY5H | 2410.08172v1 | On Evaluation of Generative Robotic Simulations | {
"content": "## Abstract\n\nAbstract Due to the difficulty of acquiring extensive real-world data, robot simulation has become crucial for parallel training and sim-to-real transfer, highlighting the importance of scalable simulated robotic tasks.\nFoundation models have demonstrated impressive capacities in autonom... | [
{
"id": "dTgnvpBfAF",
"initial_rating": 3,
"confidence": 4,
"soundness": 2,
"contribution": 2,
"presentation": 2,
"summary": "This paper proposes a framework to analyze existing simulation benchmarks for embodied AI. The authors consider 3 metrics: quality, diversity, and generalization.... | {
"rating": "3;3;5;8",
"rating_avg": 4.75,
"confidence": "4;4;3;3",
"confidence_avg": 3.5,
"soundness": "2;2;2;3",
"soundness_avg": 2.25,
"contribution": "3;2;2;3",
"contribution_avg": 2.5,
"presentation": "3;2;2;3",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:03.802407"
} | {
"id": "wXmGu4QC31",
"metareview": "The authors of this paper take on a very challenging task: analyzing the quality, diversity, and generalization of various robotics simulators. They evaluate these based on a mixture of different methods, using large language models for quality, world-model predictions for diver... | {
"decision": "Reject"
} |
s4Wm71LFK4 | 2407.20912v2 | What Are Good Positional Encodings for Directed Graphs? | {
"content": "## Abstract\n\nAbstract Positional encodings (PEs) are essential for building powerful and expressive graph neural networks and graph transformers, as they effectively capture the relative spatial relationships between nodes. Although extensive research has been devoted to PEs in undirected graphs, PEs ... | [
{
"id": "FtES17q33v",
"initial_rating": 8,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 4,
"summary": "The authors propose (i) the notion of walk profile in a directed graph for counting bidirectional walks with given numbers of forward and backward edges between t... | {
"rating": "5;5;6;8;8",
"rating_avg": 6.4,
"confidence": "4;3;3;4;4",
"confidence_avg": 3.6,
"soundness": "3;2;3;4;3",
"soundness_avg": 3,
"contribution": "3;2;2;4;3",
"contribution_avg": 2.8,
"presentation": "2;2;3;4;4",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.803252"
} | {
"id": "Nc9IWJ6ksL",
"metareview": "This paper focuses on position encodings (basically node embeddings) for directed graphs. They introduce a notion of a 'walk profile', which captures natural structures in directed graphs, show that existing embedding approaches cannot accurately represent the walk profile, and ... | {
"decision": "Accept (Poster)"
} |
s5N7p5UjgR | 2404.18988v3 | Markovian Transformers for Informative Language Modeling | {
"content": "## Abstract\n\nAbstract Chain-of-Thought (CoT) reasoning holds great promise for explaining the outputs of language models, but recent studies have highlighted significant challenges in its practical application for interpretability. We propose to address this issue via two key components: a technique t... | [
{
"id": "M9AyTQlHF1",
"initial_rating": 5,
"confidence": 2,
"soundness": 3,
"contribution": 2,
"presentation": 2,
"summary": "This paper proposes a framework where the reasoning steps are used as fixed-size states, which limits the model’s context (text bottleneck), and force the model t... | {
"rating": "5;5;6;8",
"rating_avg": 6,
"confidence": "4;2;4;4",
"confidence_avg": 3.5,
"soundness": "3;3;2;4",
"soundness_avg": 3,
"contribution": "3;2;2;4",
"contribution_avg": 2.75,
"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": "2026-01-14T22:16:03.804208"
} | {
"id": "QyW6p9bF2I",
"metareview": "This paper proposes an approach to improving the reasoning capabilities of language models by using CoT as an information bottleneck. Using a reward based on the differences between the predictions of a model conditioned on context vs not, the authors use RL techniques to train ... | {
"decision": "Reject"
} |
s5epFPdIW6 | 2410.13085v1 | MMed-RAG: Versatile Multimodal RAG System for Medical Vision Language Models | {
"content": "## Abstract\n\nAbstract Artificial Intelligence (AI) has demonstrated significant potential in healthcare, particularly in disease diagnosis and treatment planning. Recent progress in Medical Large Vision-Language Models (Med-LVLMs) has opened up new possibilities for interactive diagnostic tools. Howev... | [
{
"id": "wBlLwlVrGy",
"initial_rating": 6,
"confidence": 4,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "This work proposes MMed-RAG, a RAG system that corrects for misalignment issues for context introduced by RAG for multimodal large vision models (LVLMs). MMed-RAG... | {
"rating": "5;5;6;8",
"rating_avg": 6,
"confidence": "4;4;4;3",
"confidence_avg": 3.75,
"soundness": "3;2;3;3",
"soundness_avg": 2.75,
"contribution": "3;2;3;3",
"contribution_avg": 2.75,
"presentation": "4;3;3;4",
"presentation_avg": 3.5
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.804976"
} | {
"id": "mOmM5IieeD",
"metareview": "Key Strengths:\n1. The paper introduces RAG-PT (RAG-based preference fine-tuning) to improve alignment when using RAG with medical Large Vision Language Models (LVLMs)\n2. Thorough testing against various medical LVLMs and detailed analysis of multimodal RAG strategies\n3. Shows... | {
"decision": "Accept (Poster)"
} |
s5orchdb33 | 2409.20089v1 | Robust LLM safeguarding via refusal feature adversarial training | {
"content": "## Abstract\n\nAbstract Large language models (LLMs) are vulnerable to adversarial attacks that can elicit harmful responses. Defending against such attacks remains challenging due to the opacity of jailbreaking mechanisms and the high computational cost of training LLMs robustly. We demonstrate that ad... | [
{
"id": "WU2MdijzF4",
"initial_rating": 6,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "This paper addresses the vulnerability of large language models (LLMs) to adversarial attacks, which can elicit harmful responses by bypassing model safeguards. T... | {
"rating": "3;5;6;6",
"rating_avg": 5,
"confidence": "4;3;5;3",
"confidence_avg": 3.75,
"soundness": "3;3;3;3",
"soundness_avg": 3,
"contribution": "2;3;3;3",
"contribution_avg": 2.75,
"presentation": "2;1;3;3",
"presentation_avg": 2.25
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.805817"
} | {
"id": "rxQSNpnqWO",
"metareview": "This paper introduces a defense method for protecting LLMs against jailbreaking, based on the \"refusal feature\" observation of Arditi et al (2024). It first shows that several adversarial attacks ablate the same refusal feature direction and based on this observation design an... | {
"decision": "Accept (Poster)"
} |
s6nYndMwG7 | 2409.17357v1 | Revisiting inverse Hessian vector products for calculating influence functions | {
"content": "## Abstract\n\nAbstract Influence functions are a popular tool for attributing a model’s output to training data. The traditional approach relies on the calculation of inverse Hessian-vector products (iHVP), but the classical solver “Linear time Stochastic Second-order Algorithm” (LiSSA, Agarwal et al. ... | [
{
"id": "AEK4NrMCXM",
"initial_rating": 5,
"confidence": 3,
"soundness": 2,
"contribution": 2,
"presentation": 3,
"summary": "Influence function captures how the optimal model weights change with respect to a newly added data point and has the form of an inverse Hessian-vector product. L... | {
"rating": "3;5;5;6",
"rating_avg": 4.75,
"confidence": "4;3;3;4",
"confidence_avg": 3.5,
"soundness": "2;3;2;3",
"soundness_avg": 2.5,
"contribution": "2;2;2;3",
"contribution_avg": 2.25,
"presentation": "1;3;3;3",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:03.806577"
} | {
"id": "1E5O9q4nPI",
"metareview": "This paper analyzes the convergence for LiSSA algorithm (used to compute inverse Hessian-vector products, which are useful for computing influence functions). The results show that a large (but mild) batch is required for convergence, and the optimal hyperparameters for the algo... | {
"decision": "Reject"
} |
s6q6zX45F8 | 2402.18888v3 | Uncertainty-Based Extensible Codebook for Discrete Federated Learning in Heterogeneous Data Silos | {
"content": "## Abstract\n\nAbstract Federated learning (FL), aimed at leveraging vast distributed datasets, confronts a crucial challenge: the heterogeneity of data across different silos. While previous studies have explored discrete representations to enhance model generalization across minor distributional shift... | [
{
"id": "4GhCpPUFQZ",
"initial_rating": 5,
"confidence": 4,
"soundness": 2,
"contribution": 2,
"presentation": 2,
"summary": "The paper addresses the challenge of data heterogeneity in Federated Learning (FL) by proposing a novel method called Uncertainty-Based Extensible-Codebook Federa... | {
"rating": "5;5;6",
"rating_avg": 5.333333333333333,
"confidence": "3;4;4",
"confidence_avg": 3.6666666666666665,
"soundness": "2;2;3",
"soundness_avg": 2.3333333333333335,
"contribution": "2;2;3",
"contribution_avg": 2.3333333333333335,
"presentation": "3;2;3",
"presentation_avg": 2.66666666666666... | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:03.807277"
} | {
"id": "1Md3VjKsLc",
"metareview": "Post-rebuttal, the majority of the reviewers still judged the contribution of the paper as fair. I appreciate the efforts the authors have put in their rebuttal, and the updates in the draft, very substantial and helpful in the appendices, but in the end the paper unfortunately ... | {
"decision": "Reject"
} |
s7lzZpAW7T | 2411.05361v1 | Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks | {
"content": "## Abstract\n\nAbstract Multimodal foundation models, such as Gemini and ChatGPT, have revolutionized human-machine interactions by seamlessly integrating various forms of data.\nDeveloping a universal spoken language model that comprehends a wide range of natural language instructions is critical for b... | [
{
"id": "Br2cGBN3I1",
"initial_rating": 6,
"confidence": 4,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "This paper builds on the prior work, Dynamic-Superb, by expanding its scope as an audio LLM benchmark to include a broader range of tasks centered on audio and sp... | {
"rating": "6;6;8;8",
"rating_avg": 7,
"confidence": "3;4;3;3",
"confidence_avg": 3.25,
"soundness": "3;3;3;3",
"soundness_avg": 3,
"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": "2026-01-14T22:16:03.808119"
} | {
"id": "ioKhQlOIbV",
"metareview": "Dynamic-SUPERB Phase-2 is the largest and most comprehensive benchmark for instruction-based universal speech models, covering a wide range of tasks across speech, music, and audio with natural language instructions. It tackles the significant problem of the lack of audio-text ... | {
"decision": "Accept (Poster)"
} |
s7vwXDsVYa | 2405.14857v3 | Conditional Diffusion on Web-Scale Image Pairs leads to Diverse Image Variations | {
"content": "## Abstract\n\nAbstract Generating image variations, where a model produces variations of an input image while preserving the semantic context has gained increasing attention. Current image variation techniques involve adapting a text-to-image model to reconstruct an input image conditioned on the same ... | [
{
"id": "5oWYSj9Z04",
"initial_rating": 3,
"confidence": 5,
"soundness": 3,
"contribution": 2,
"presentation": 2,
"summary": "This paper focuses on the task of generating image variations and proposes the Semantica model. Unlike traditional diffusion model training, Semantica is trained ... | {
"rating": "3;3;5;8",
"rating_avg": 4.75,
"confidence": "3;5;3;3",
"confidence_avg": 3.5,
"soundness": "2;3;2;3",
"soundness_avg": 2.5,
"contribution": "2;2;2;3",
"contribution_avg": 2.25,
"presentation": "2;2;4;3",
"presentation_avg": 2.75
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:03.808913"
} | {
"id": "BB9RvuK3Ii",
"metareview": "This work targets on generating image variations by training the model with web-scale image pairs. The idea of leveraging the semantic relations of image pairs and simplicity of method are appreciated. However, several main concerns raised by reviewers include the unclear motiva... | {
"decision": "Reject"
} |
sAxdIJ4l6z | 2410.07113v1 | Personalized Visual Instruction Tuning | {
"content": "## Abstract\n\nAbstract Recent advancements in multimodal large language models (MLLMs) have demonstrated significant progress; however, these models exhibit a notable limitation, which we refer to as “face blindness”. Specifically, they can engage in general conversations but fail to conduct personaliz... | [
{
"id": "hqXWFoMfy9",
"initial_rating": 5,
"confidence": 4,
"soundness": 3,
"contribution": 2,
"presentation": 2,
"summary": "This paper tackles personalization issue in MLLM. Current MLLM struggle to recognize and engage with specific people in images, making them less useful for person... | {
"rating": "5;5;6;6",
"rating_avg": 5.5,
"confidence": "4;4;5;4",
"confidence_avg": 4.25,
"soundness": "3;3;2;3",
"soundness_avg": 2.75,
"contribution": "2;3;2;3",
"contribution_avg": 2.5,
"presentation": "3;3;2;3",
"presentation_avg": 2.75
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.809855"
} | {
"id": "huTxRTYY6K",
"metareview": "This paper addresses the \"face blindness\" limitation in multimodal large language models (MLLMs) by introducing Personalized Visual Instruction Tuning (PVIT), a novel framework for enabling personalized dialogues about specific individuals in images. The key scientific contrib... | {
"decision": "Accept (Poster)"
} |
sBJIVQvJqN | 2410.10879v1 | Enhancing Vision-Language Model Pre-training with Image-text Pair Pruning Based on Word Frequency | {
"content": "## Abstract\n\nAbstract We propose W ord- F requency-based Image-Text P air P runing (WFPP), a novel data pruning method that improves the efficiency of VLMs.\nUnlike MetaCLIP, our method does not need metadata for pruning, but selects text-image pairs to prune based on the content of the text. Specific... | [
{
"id": "hrCfulHSlI",
"initial_rating": 3,
"confidence": 3,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "This paper introduces Word-Frequency-based Image-Text Pair Pruning (WFPP), a method to enhance the pretraining of Vision-Language Models (VLMs) by pruning the tra... | {
"rating": "3;5;5;8",
"rating_avg": 5.25,
"confidence": "3;4;3;3",
"confidence_avg": 3.25,
"soundness": "3;3;2;3",
"soundness_avg": 2.75,
"contribution": "2;2;2;3",
"contribution_avg": 2.25,
"presentation": "3;2;3;3",
"presentation_avg": 2.75
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:03.810756"
} | {
"id": "zwWfIW1n9I",
"metareview": "This paper proposes WFPP, a word-frequency-based pruning method for improving CLIP pre-training by removing image-text pairs containing high-frequency words. While the paper demonstrates some empirical improvements on CC3M/CC12M datasets and offers a simple, practical approach, ... | {
"decision": "Reject"
} |
sBbarJBdkn | 2406.11785v2 | CELL your Model: Contrastive Explanations for Large Language Models | {
"content": "## Abstract\n\nAbstract The advent of black-box deep neural network classification models has sparked the need to explain their decisions. However, in the case of generative AI, such as large language models (LLMs), there is no class prediction to explain. Rather, one can ask why an LLM output a particu... | [
{
"id": "zogRQDYSpM",
"initial_rating": 3,
"confidence": 4,
"soundness": 2,
"contribution": 2,
"presentation": 2,
"summary": "This paper proposes a contrastive explanation method for large language models that require simply black-box/query access.The method requires a scoring function a... | {
"rating": "3;3;3;5;5",
"rating_avg": 3.8,
"confidence": "3;5;4;4;4",
"confidence_avg": 4,
"soundness": "2;2;2;3;2",
"soundness_avg": 2.2,
"contribution": "2;2;2;2;2",
"contribution_avg": 2,
"presentation": "2;2;2;2;3",
"presentation_avg": 2.2
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:03.811509"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
sCew1tR6No | 2402.03664v3 | Partial Gromov-Wasserstein Metric | {
"content": "## Abstract\n\nAbstract The Gromov-Wasserstein (GW) distance has gained increasing interest in the machine learning community in recent years, as it allows for the comparison of measures in different metric spaces. To overcome the limitations imposed by the equal mass requirements of the classical GW pr... | [
{
"id": "XWKv3PHRKh",
"initial_rating": 6,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 2,
"summary": "This paper improves unbalanced approaches to the optimal transport (OT) problem. \nThe objective is to compare probability measures between different metric space... | {
"rating": "5;5;6;6;8",
"rating_avg": 6,
"confidence": "4;4;3;4;4",
"confidence_avg": 3.8,
"soundness": "3;2;3;3;4",
"soundness_avg": 3,
"contribution": "2;2;2;3;3",
"contribution_avg": 2.4,
"presentation": "3;2;3;2;3",
"presentation_avg": 2.6
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.812676"
} | {
"id": "0vkmRzLkBZ",
"metareview": "The paper is a nice contribution, introducing the Partial Gromov-Wasserstein (PGW) metric, which addresses limitations in existing Gromov-Wasserstein formulations for unbalanced settings. The authors provide rigorous theoretical guarantees, such as proving that PGW is a well-def... | {
"decision": "Accept (Poster)"
} |
sClhxLqfnP | 2410.08181v1 | RGM: Reconstructing High-fidelity 3D Car Assets with Relightable 3D-GS Generative Model from a Single Image | {
"content": "## Abstract\n\nAbstract The generation of high-quality 3D car assets is essential for various applications, including video games, autonomous driving, and virtual reality. Current 3D generation methods utilizing NeRF or 3D-GS as representations for 3D objects, generate a Lambertian object under fixed li... | [
{
"id": "bktsKIVYY3",
"initial_rating": 3,
"confidence": 3,
"soundness": 1,
"contribution": 3,
"presentation": 2,
"summary": "The main contribution of the paper is the introduction of Relightable 3D-GS Generative Model (RGM), a novel framework for automatically generating high-fidelity 3... | {
"rating": "3;3;5;6",
"rating_avg": 4.25,
"confidence": "4;3;4;4",
"confidence_avg": 3.75,
"soundness": "3;1;3;3",
"soundness_avg": 2.5,
"contribution": "2;3;2;2",
"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": "2026-01-14T22:16:03.813984"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
sGqd1tF8P8 | 2409.08813v1 | Your Weak LLM is Secretly a Strong Teacher for Alignment | {
"content": "## Abstract\n\nAbstract The burgeoning capabilities of large language models (LLMs) have underscored the need for alignment to ensure these models act in accordance with human values and intentions. Existing alignment frameworks present constraints either in the form of expensive human effort or high co... | [
{
"id": "DQ640QzSs7",
"initial_rating": 5,
"confidence": 4,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "This paper concentrates on the critical research topic of large language model (LLM) alignment, especially related to super-alignment. Considering existing mainst... | {
"rating": "3;5;5;6;8",
"rating_avg": 5.4,
"confidence": "4;4;4;3;4",
"confidence_avg": 3.8,
"soundness": "2;3;3;3;4",
"soundness_avg": 3,
"contribution": "2;2;2;3;3",
"contribution_avg": 2.4,
"presentation": "3;3;4;3;4",
"presentation_avg": 3.4
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.814710"
} | {
"id": "KCLSSFrPwc",
"metareview": "This work focuses on LLM alignment and proposes a method that employs a semi-supervised framework. In this framework, a weak LLM, trained on a small labeled dataset, provides preference feedback for a larger unlabeled dataset. Experiments show its effectiveness in alignment task... | {
"decision": "Accept (Poster)"
} |
sHAvMp5J4R | 2410.06166v1 | Temporal Reasoning Transfer from Text to Video | {
"content": "## Abstract\n\nAbstract Video Large Language Models (Video LLMs) have shown promising capabilities in video comprehension, yet they struggle with tracking temporal changes and reasoning about temporal relationships.\nWhile previous research attributed this limitation to the ineffective temporal encoding... | [
{
"id": "uxrkPUBvKU",
"initial_rating": 5,
"confidence": 3,
"soundness": 2,
"contribution": 2,
"presentation": 2,
"summary": "This paper addresses the limitations of video LLMs in temporal reasoning, which involves tracking objects or events in a sequence of video frames. The authors dis... | {
"rating": "5;5;5;5;6",
"rating_avg": 5.2,
"confidence": "4;4;4;3;4",
"confidence_avg": 3.8,
"soundness": "3;2;3;2;3",
"soundness_avg": 2.6,
"contribution": "3;2;2;2;3",
"contribution_avg": 2.4,
"presentation": "4;3;3;2;3",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.815532"
} | {
"id": "rOepzlLa62",
"metareview": "This work presented an innovative perspective about the temporal reasoning ability for multimodal LLMs. The authors found that the drawback of temporal reasoning ability mainly stems from the weak performance of LLMs in this aspect. Starting from this observation, the authors cu... | {
"decision": "Accept (Poster)"
} |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.