paper_id string | arxiv_id string | title string | markdown dict | reviews list | scores dict | metadata dict | meta_review dict | decision dict |
|---|---|---|---|---|---|---|---|---|
9JE3HogPCw | 2406.09079v2 | Hadamard Representations: Augmenting Hyperbolic Tangents in RL | {
"content": "## Abstract\n\nAbstract Activation functions are one of the key components of a deep neural network. The most commonly used activation functions can be classed into the category of continuously differentiable (e.g. tanh) and linear-unit functions (e.g. ReLU), both having their own strengths and drawback... | [
{
"id": "ZYeZVb6YzA",
"initial_rating": 3,
"confidence": 4,
"soundness": 2,
"contribution": 2,
"presentation": 3,
"summary": "This paper studies the dying neuron problem with tanh-type activations. The paper's first result is that bounded activation functions like tanh perform worse than... | {
"rating": "3;3;5;5",
"rating_avg": 4,
"confidence": "4;4;4;4",
"confidence_avg": 4,
"soundness": "2;2;3;2",
"soundness_avg": 2.25,
"contribution": "2;2;3;2",
"contribution_avg": 2.25,
"presentation": "3;3;3;4",
"presentation_avg": 3.25
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:01.065743"
} | {
"id": "4mFyJC5W0Q",
"metareview": "In this paper, the authors proposes a novel activation function based on Hadamard representations (HR) to address limitations of traditional activation functions in reinforcement learning, such as vanishing gradients and dying neurons. The authors demonstrate improved learning e... | {
"decision": "Reject"
} |
9KatbAXLAq | 2404.05350v1 | Certified PEFTSmoothing: Parameter-Efficient Fine-Tuning with Randomized Smoothing | {
"content": "## Abstract\n\nAbstract Randomized smoothing is the primary certified robustness method for accessing the robustness of deep learning models to adversarial perturbations in the l 2 subscript 𝑙 2 l_{2} italic_l start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT -norm, by adding isotropic Gaussian noise to the inpu... | [
{
"id": "jSURj6AcoA",
"initial_rating": 6,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 2,
"summary": "In this paper, the authors propose a PEFT-inspired method for adapting pre-trained base models to produce robust classification performance under noise-augmented ... | {
"rating": "3;3;3;6",
"rating_avg": 3.75,
"confidence": "4;3;3;3",
"confidence_avg": 3.25,
"soundness": "2;2;2;3",
"soundness_avg": 2.25,
"contribution": "1;1;2;3",
"contribution_avg": 1.75,
"presentation": "3;2;2;2",
"presentation_avg": 2.25
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:01.066461"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
9LdJDU7E91 | 2405.17238v2 | LLM-Assisted Static Analysis for Detecting Security Vulnerabilities | {
"content": "## Abstract\n\nAbstract Software is prone to security vulnerabilities.\nProgram analysis tools to detect them have limited effectiveness in practice due to their reliance on human labeled specifications.\nLarge language models (or LLMs) have shown impressive code generation capabilities but they cannot ... | [
{
"id": "lwr7PR9V3m",
"initial_rating": 3,
"confidence": 4,
"soundness": 2,
"contribution": 2,
"presentation": 3,
"summary": "This paper presents a hybrid approach that combines static analysis (CodeQL) with LLMs to detect vulnerabilities. Specifically, IRIS uses LLMs to find taint speci... | {
"rating": "3;6;6;8",
"rating_avg": 5.75,
"confidence": "4;4;4;4",
"confidence_avg": 4,
"soundness": "2;3;2;3",
"soundness_avg": 2.5,
"contribution": "2;3;2;3",
"contribution_avg": 2.5,
"presentation": "3;4;2;3",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.067254"
} | {
"id": "if8nqpSnwp",
"metareview": "This paper proposes IRIS, a neuro-symbolic approach for security vulnerability detection that integrates large language models with static analysis. Overall, it is a borderline submission with mixed reviews. While most reviewers provide positive feedback, reviewer EvMB holds a d... | {
"decision": "Accept (Poster)"
} |
9M5georQ9T | 2410.14581v1 | Optimizing Attention with Mirror Descent: Generalized Max-Margin Token Selection | {
"content": "## Abstract\n\nAbstract Attention mechanisms have revolutionized several domains of artificial intelligence, such as natural language processing and computer vision, by enabling models to selectively focus on relevant parts of the input data. While recent work has characterized the optimization dynamics... | [
{
"id": "2iWOd2f9BP",
"initial_rating": 6,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "The paper introduces a novel approach to optimizing attention mechanisms using Mirror Descent (MD), specifically focusing on a generalized max-margin token select... | {
"rating": "3;3;6;6",
"rating_avg": 4.5,
"confidence": "3;4;3;3",
"confidence_avg": 3.25,
"soundness": "2;2;3;3",
"soundness_avg": 2.5,
"contribution": "2;2;3;3",
"contribution_avg": 2.5,
"presentation": "2;4;3;3",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:01.068188"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
9NfHbWKqMF | 2411.06390v2 | SplatFormer: Point Transformer for Robust 3D Gaussian Splatting | {
"content": "## Abstract\n\nAbstract 3D Gaussian Splatting (3DGS) has recently transformed photorealistic reconstruction, achieving high visual fidelity and real-time performance. However, rendering quality significantly deteriorates when test views deviate from the camera angles used during training, posing a major... | [
{
"id": "FED4lhpknc",
"initial_rating": 5,
"confidence": 5,
"soundness": 2,
"contribution": 3,
"presentation": 3,
"summary": "The paper introduces SplatFormer, a point transformer model for refining 3D Gaussian Splatting (3DGS) representations under out-of-distribution (OOD) view conditi... | {
"rating": "5;6;6;8",
"rating_avg": 6.25,
"confidence": "5;4;3;3",
"confidence_avg": 3.75,
"soundness": "2;4;3;3",
"soundness_avg": 3,
"contribution": "3;3;3;3",
"contribution_avg": 3,
"presentation": "3;3;3;3",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Spotlight",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.069205"
} | {
"id": "qvsowSaCVL",
"metareview": "This paper receives unanimous positive ratings of 6,8,8,8. The AC follows the recommendations of the reviewers to accept the paper. The reviewers comment that the method introduced by the paper is novel and the task is an important direction for rendering unseen test views which... | {
"decision": "Accept (Spotlight)"
} |
9OMvtboTJg | 2410.13213v1 | LLMOPT: Learning to Define and Solve General Optimization Problems from Scratch | {
"content": "## Abstract\n\nAbstract Optimization problems are prevalent across various scenarios. Formulating and then solving optimization problems described by natural language often requires highly specialized human expertise, which could block the widespread application of optimization-based decision making. To... | [
{
"id": "wKlQN4dkW1",
"initial_rating": 6,
"confidence": 4,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "The authors of this paper introduced **LLMOPT**, a novel learning-based framework designed to enhance large language models' (LLMs) capability to define and solve... | {
"rating": "3;5;5;6",
"rating_avg": 4.75,
"confidence": "4;3;3;4",
"confidence_avg": 3.5,
"soundness": "2;3;3;3",
"soundness_avg": 2.75,
"contribution": "2;2;2;2",
"contribution_avg": 2,
"presentation": "2;3;2;3",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.070012"
} | {
"id": "Jk35XXQv3k",
"metareview": "This paper introduces LLMOPT, which fine-tunes LLMs to enhance their ability to model optimization problems from natural language descriptions. Specifically, the fine-tuning process consists of two stages: (1) the supervised fine-tuning (SFT) stage, where the LLM is trained usin... | {
"decision": "Accept (Poster)"
} |
9OfKxKoYNw | 2410.05694v1 | DiffusionGuard: A Robust Defense Against Malicious Diffusion-based Image Editing | {
"content": "## Abstract\n\nAbstract Recent advances in diffusion models have introduced a new era of text-guided image manipulation, enabling users to create realistic edited images with simple textual prompts.\nHowever, there is significant concern about the potential misuse of these methods, especially in creatin... | [
{
"id": "xkdxmhiKmp",
"initial_rating": 3,
"confidence": 2,
"soundness": 2,
"contribution": 2,
"presentation": 3,
"summary": "The paper proposes DiffusionGuard, an image-cloaking algorithm to defend against malicious diffusion-based text-guided inpainting. Compared to previous works, it ... | {
"rating": "3;5;5;6",
"rating_avg": 4.75,
"confidence": "2;3;3;4",
"confidence_avg": 3,
"soundness": "2;3;3;3",
"soundness_avg": 2.75,
"contribution": "2;3;2;3",
"contribution_avg": 2.5,
"presentation": "3;3;3;3",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.070795"
} | {
"id": "M2RzFGCfOQ",
"metareview": "This work introduces a novel adversarial noise-based defense method designed to protect images from unauthorized edits by diffusion-based image editing techniques. The authors propose an objective that targets the early stages of the diffusion process to enhance attack performan... | {
"decision": "Accept (Poster)"
} |
9Orm76dUuT | 2402.08577v1 | Test-Time Backdoor Attacks on Multimodal Large Language Models | {
"content": "## Abstract\n\nAbstract Backdoor attacks are commonly executed by contaminating training data, such that a trigger can activate predetermined harmful effects during the test phase. In this work, we present AnyDoor , a test-time backdoor attack against multimodal large language models (MLLMs), which invo... | [
{
"id": "WjZPIyk8N7",
"initial_rating": 5,
"confidence": 3,
"soundness": 2,
"contribution": 2,
"presentation": 3,
"summary": "This paper explores the possibility of implementing a backdoor attack during the testing phase. The paper proposes a type of backdoor attack called ``AnyDoor``, w... | {
"rating": "3;5;5;5",
"rating_avg": 4.5,
"confidence": "4;4;4;3",
"confidence_avg": 3.75,
"soundness": "2;3;3;2",
"soundness_avg": 2.5,
"contribution": "2;2;3;2",
"contribution_avg": 2.25,
"presentation": "3;3;2;3",
"presentation_avg": 2.75
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:01.071578"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
9P5I9zTUAd | 2404.18410v1 | Mixture-of-Instructions: Aligning Large Language Models via Mixture Prompting | {
"content": "## Abstract\n\nAbstract With the proliferation of large language models (LLMs), the comprehensive alignment of such models across multiple tasks has emerged as a critical area of research. Existing alignment methodologies primarily address single task, such as multi-turn dialogue, coding, mathematical p... | [
{
"id": "EPnbrpRqTH",
"initial_rating": 6,
"confidence": 4,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "The paper introduces a technique known as Mixture-of-Instructions (MoI) for improving the alignment efficiency of large language models (LLMs) across multiple tas... | {
"rating": "3;3;5;6;6",
"rating_avg": 4.6,
"confidence": "4;4;4;4;4",
"confidence_avg": 4,
"soundness": "2;3;4;3;3",
"soundness_avg": 3,
"contribution": "2;2;3;3;2",
"contribution_avg": 2.4,
"presentation": "2;3;4;3;3",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:01.072255"
} | {
"id": "qN0CVm9FXu",
"metareview": "This paper proposes an approach, Mixture-of-Instructions (MoI), that aims to align a model with several tasks simultaneously. The approach is based on instruction packing with diverse system prompts to represent and improve multiple downstream tasks during supervised fine-tuning... | {
"decision": "Reject"
} |
9Q9KXUTjmd | 2408.14144v2 | Neighborhood and Global Perturbations Supported SAM in Federated Learning: From Local Tweaks To Global Awareness | {
"content": "## Abstract\n\nAbstract Federated Learning (FL) can be coordinated under the orchestration of a central server to collaboratively build a privacy-preserving model without the need for data exchange.\nHowever, participant data heterogeneity leads to local optima divergence, subsequently affecting converg... | [
{
"id": "q2PPxJPinX",
"initial_rating": 3,
"confidence": 3,
"soundness": 3,
"contribution": 2,
"presentation": 2,
"summary": "This paper presents FedTOGA, a Federated Learning (FL) algorithm designed to prevent the heterogeneity of client data cause the global model to converge to a sha... | {
"rating": "3;3;5;5",
"rating_avg": 4,
"confidence": "4;3;4;3",
"confidence_avg": 3.5,
"soundness": "2;3;3;2",
"soundness_avg": 2.5,
"contribution": "2;2;3;3",
"contribution_avg": 2.5,
"presentation": "1;2;2;2",
"presentation_avg": 1.75
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:01.072928"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
9Qptgv0Eyw | 2410.17377v1 | PtychoFormer: A Transformer-based Model for Ptychographic Phase Retrieval | {
"content": "## Abstract\n\nAbstract Ptychography is a computational method of microscopy that recovers high-resolution transmission images of samples from a series of diffraction patterns. While conventional phase retrieval algorithms can iteratively recover the images, they require oversampled diffraction patterns... | [
{
"id": "ZFzHPEZBiR",
"initial_rating": 5,
"confidence": 3,
"soundness": 2,
"contribution": 2,
"presentation": 3,
"summary": "This paper studies ptychographic phase retrieval and proposes a transformer-based model for recovering phase and amplitude from a series of diffraction patterns. ... | {
"rating": "1;3;3;5;5",
"rating_avg": 3.4,
"confidence": "4;4;4;4;3",
"confidence_avg": 3.8,
"soundness": "1;2;2;2;2",
"soundness_avg": 1.8,
"contribution": "1;2;2;1;2",
"contribution_avg": 1.6,
"presentation": "2;1;3;2;3",
"presentation_avg": 2.2
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:01.073634"
} | {
"id": "5lrQYtjkpg",
"metareview": "The paper proposes Ptychoformer, a hirerarchical transformer-based method for solving the ptychographic phase retrieval problem. \n\nThe algorithm runs quickly and outperforms earlier (but somewhat outdated) neural-network based methods for phase retrieval. \n\nThe reviewers rai... | {
"decision": "Reject"
} |
9RCT0ngvZP | 2410.14208v1 | Montessori-Instruct: Generate Influential Training Data Tailored for Student Learning | {
"content": "## Abstract\n\nAbstract Synthetic data has been widely used to train large language models, but their generative nature inevitably introduces noisy, non-informative, and misleading learning signals. In this paper, we propose Montessori-Instruct , a novel data synthesis framework that tailors the data sy... | [
{
"id": "oAafNHmG2z",
"initial_rating": 6,
"confidence": 3,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "This paper introduces Montessori-Instruct, a new framework for generating synthetic training data to enhance the learning of student language models. The authors ... | {
"rating": "3;5;6;6",
"rating_avg": 5,
"confidence": "3;3;3;4",
"confidence_avg": 3.25,
"soundness": "3;3;3;3",
"soundness_avg": 3,
"contribution": "2;2;3;3",
"contribution_avg": 2.5,
"presentation": "3;3;3;3",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.074566"
} | {
"id": "fkGMsOQyaA",
"metareview": "This paper proposes to use influence function to select training examples in the iterative instruction tuning process. Despite increasing the computation cost due to evaluating the influence function, it can outperform other iterative improvement methods and the observations are... | {
"decision": "Accept (Poster)"
} |
9RFocgIccP | 2411.04713v1 | Multi-Reward as Condition for Instruction-based Image Editing | {
"content": "## Abstract\n\nAbstract High-quality training triplets (instruction, original image, edited image) are essential for instruction-based image editing.\nPredominant training datasets (e.g., InsPix2Pix) are created using text-to-image generative models (e.g., Stable Diffusion, DALL-E) which are not trained... | [
{
"id": "cGxgOsUrcz",
"initial_rating": 6,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "The authors introduce a high-quality editing reward dataset to address the problem of editing effectiveness due to training data quality issues with instruction-b... | {
"rating": "5;5;6;6",
"rating_avg": 5.5,
"confidence": "5;3;4;4",
"confidence_avg": 4,
"soundness": "3;3;3;3",
"soundness_avg": 3,
"contribution": "2;4;3;3",
"contribution_avg": 3,
"presentation": "2;3;3;3",
"presentation_avg": 2.75
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.075267"
} | {
"id": "MDe4jJSHtg",
"metareview": "This paper aims to use multi-view reward data as an additional condition to address the problem of editing effectiveness caused by the issues of training data quality. To achieve this, the authors introduce a high-quality editing reward dataset and propose a benchmark to evaluat... | {
"decision": "Accept (Poster)"
} |
9RcofuNF5p | 2403.12529v3 | Contextualized Messages Boost Graph Representations | {
"content": "## Abstract\n\nAbstract Graph neural networks (GNNs) have gained significant attention in recent years for their ability to process data that may be represented as graphs. This has prompted several studies to explore their representational capability based on the graph isomorphism task. These works inhe... | [
{
"id": "wFQBIibEXu",
"initial_rating": 3,
"confidence": 2,
"soundness": 2,
"contribution": 2,
"presentation": 2,
"summary": "This paper extends PNA, which considers uncountable node features, by incorporating anisotropic and dynamic. The difference between PNA and the proposed SIR-GCN i... | {
"rating": "3;6;6",
"rating_avg": 5,
"confidence": "2;3;4",
"confidence_avg": 3,
"soundness": "2;3;3",
"soundness_avg": 2.6666666666666665,
"contribution": "2;3;3",
"contribution_avg": 2.6666666666666665,
"presentation": "2;3;2",
"presentation_avg": 2.3333333333333335
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:01.076075"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
9SYczU3Qgm | 2408.14608v1 | Meta Flow Matching: Integrating Vector Fields on the Wasserstein Manifold | {
"content": "## Abstract\n\nAbstract Numerous biological and physical processes can be modeled as systems of interacting entities evolving continuously over time, e.g. the dynamics of communicating cells or physical particles.\nLearning the dynamics of such systems is essential for predicting the temporal evolution ... | [
{
"id": "cRhPU4lWzR",
"initial_rating": 8,
"confidence": 4,
"soundness": 3,
"contribution": 4,
"presentation": 4,
"summary": "This paper introduces Meta Flow Matching (MFM), a novel approach for modeling the evolution of systems consisting of interacting samples/populations. Unlike previ... | {
"rating": "5;5;5;8",
"rating_avg": 5.75,
"confidence": "3;4;2;4",
"confidence_avg": 3.25,
"soundness": "3;2;2;3",
"soundness_avg": 2.5,
"contribution": "3;3;3;3",
"contribution_avg": 3,
"presentation": "2;3;1;3",
"presentation_avg": 2.25
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.076810"
} | {
"id": "PQzjLxPyxe",
"metareview": "The paper proposes a 'meta flow matching approach' for modelling a family of distributions, conditioned on the population index i; \ninstead of previous approaches (conditional generative flow matching) instead of embedding a sequence i the meta-flow matching approach embeds to ... | {
"decision": "Accept (Poster)"
} |
9SmukfhJoF | 2410.01647v1 | 3DGS-Det: Empower 3D Gaussian Splatting with Boundary Guidance and Box-Focused Sampling for 3D Object Detection | {
"content": "## Abstract\n\nAbstract Neural Radiance Fields (NeRF) are widely used for novel-view synthesis and have been adapted for 3D Object Detection (3DOD), offering a promising approach to 3D object detection through view-synthesis representation. However, NeRF faces inherent limitations: (i) It has limited re... | [
{
"id": "pE551tRQXA",
"initial_rating": 6,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "The paper proposes to use 3D Gaussian Splatting (3DGS) as the representation to do 3D object detection. To make 3DGS works well for the 3D object detection task, ... | {
"rating": "5;5;5;6",
"rating_avg": 5.25,
"confidence": "3;4;4;3",
"confidence_avg": 3.5,
"soundness": "2;3;2;3",
"soundness_avg": 2.5,
"contribution": "2;2;3;3",
"contribution_avg": 2.5,
"presentation": "2;3;3;3",
"presentation_avg": 2.75
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:01.077755"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
9SvRqu21m7 | 2410.23274v1 | Multi-Student Diffusion Distillation for Better One-Step Generators | {
"content": "## Abstract\n\nAbstract Diffusion models achieve high-quality sample generation at the cost of a lengthy multistep inference procedure.\nTo overcome this, diffusion distillation techniques produce student generators capable of matching or surpassing the teacher in a single step.\nHowever, the student mo... | [
{
"id": "wPsNgEgaib",
"initial_rating": 3,
"confidence": 4,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "In this work authors propose a way to distill a pre-trained diffusion model into multiple student where each student is specialized for sub-domain or specific par... | {
"rating": "3;3;3;6",
"rating_avg": 3.75,
"confidence": "3;4;4;5",
"confidence_avg": 4,
"soundness": "2;2;3;4",
"soundness_avg": 2.75,
"contribution": "2;2;2;3",
"contribution_avg": 2.25,
"presentation": "2;2;3;3",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:01.078693"
} | {
"id": "oPfthbq27d",
"metareview": "This paper presents a method for single step diffusion model distillation by turning the student to a MoE style conditioned on the data partition. It has received mixed reviews -- reviewer igyA is overall positive about the contributions, however other three reviewers challenged... | {
"decision": "Reject"
} |
9TClCDZXeh | 2406.14995v2 | Differentiable and Learnable Wireless Simulation with Geometric Transformers | {
"content": "## Abstract\n\nAbstract Modelling the propagation of electromagnetic wireless signals is critical for designing modern communication systems. Wireless ray tracing simulators model signal propagation based on the 3D geometry and other scene parameters, but\ntheir accuracy is fundamentally limited by unde... | [
{
"id": "rGC4bBEBFU",
"initial_rating": 6,
"confidence": 2,
"soundness": 3,
"contribution": 3,
"presentation": 2,
"summary": "The paper presents a learnable approach to tackle the problem of indoor wireless simulation. The proposed architecture is based on a Geometric Algebra Transformer... | {
"rating": "5;6;6;6",
"rating_avg": 5.75,
"confidence": "3;4;4;2",
"confidence_avg": 3.25,
"soundness": "3;4;3;3",
"soundness_avg": 3.25,
"contribution": "2;3;3;3",
"contribution_avg": 2.75,
"presentation": "2;3;3;2",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.079586"
} | {
"id": "smRFxHp22T",
"metareview": "In this paper, the authors present a novel geometric transformer model for wireless simulation. The reviewers are generally positive and recognize the valuable contribution of the proposed transformer model, particularly its application to wireless indoor modeling. This work has... | {
"decision": "Accept (Poster)"
} |
9TL99KnTv5 | 2402.13037v2 | Align Your Intents: Offline Imitation Learning via Optimal Transport | {
"content": "## Abstract\n\nAbstract Offline Reinforcement Learning (RL) addresses the problem of sequential decision-making by learning optimal policy through pre-collected data, without interacting with the environment.\nAs yet, it has remained somewhat impractical, because one rarely knows the reward explicitly a... | [
{
"id": "qVqEi4owmI",
"initial_rating": 5,
"confidence": 3,
"soundness": 3,
"contribution": 2,
"presentation": 1,
"summary": "This paper considers learning from offline data in settings where reward may be difficult to specify, but one (or multiple) expert trajectories demonstrating the ... | {
"rating": "3;5;5;6",
"rating_avg": 4.75,
"confidence": "4;4;3;3",
"confidence_avg": 3.5,
"soundness": "2;3;3;3",
"soundness_avg": 2.75,
"contribution": "2;2;2;3",
"contribution_avg": 2.25,
"presentation": "2;2;1;3",
"presentation_avg": 2
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:01.080265"
} | {
"id": "xP6n93R1kQ",
"metareview": "This paper proposes a method for the offline RL setting where rewards are difficult to specify but one (or multiple) expert trajectories demonstrating the behavior may be found. The proposed method computes rewards by compares the optimal transport distance (computed via ICVF) b... | {
"decision": "Reject"
} |
9VGTk2NYjF | 2409.07398v1 | The Complexity of Two-Team Polymatrix Games with Independent Adversaries | {
"content": "## 1 Introduction\n\nGame theory is a fundamental tool to encode strategic agent interactions and has found many applications in the modern AI landscape such as Generative Adversarial Networks [], obtaining agents with expert level play in multiplayer games such as Starcraft and Quake III [, ] and super... | [
{
"id": "P9JTVKBHnU",
"initial_rating": 8,
"confidence": 4,
"soundness": 4,
"contribution": 3,
"presentation": 3,
"summary": "In this work, the authors investigate the computational complexity of computing a Nash equilibrium in two-team zero-sum polymatrix games where one team consists o... | {
"rating": "8;8;8",
"rating_avg": 8,
"confidence": "3;3;4",
"confidence_avg": 3.3333333333333335,
"soundness": "4;4;4",
"soundness_avg": 4,
"contribution": "3;3;3",
"contribution_avg": 3,
"presentation": "4;4;3",
"presentation_avg": 3.6666666666666665
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Oral",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.080918"
} | {
"id": "AXSf2eOPmA",
"metareview": "This paper looks at the complexity of computing teams equilibria in game.\n\nIt is a good theoretic contribution that the reviewers and I enjoyed reading. It is well written and rather insightful.\n\nHappy to recommend acceptance.",
"additional_comments": "All positive reviews... | {
"decision": "Accept (Oral)"
} |
9W6Z9IeLzc | 2410.16670v1 | CoPS: Empowering LLM Agents with Provable Cross-Task Experience Sharing | {
"content": "## Abstract\n\nAbstract Sequential reasoning in agent systems has been significantly advanced by large language models (LLMs), yet existing approaches face limitations. Reflection-driven reasoning relies solely on knowledge in pretrained models, limiting performance in novel scenarios, while experience-... | [
{
"id": "Sqb13zx4xk",
"initial_rating": 5,
"confidence": 3,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "The authors propose COPS, a method which utilizes an offline demonstration dataset of “experiences” and study how to use these experiences to solve downstream emb... | {
"rating": "5;5;5;5",
"rating_avg": 5,
"confidence": "3;3;4;3",
"confidence_avg": 3.25,
"soundness": "3;2;3;3",
"soundness_avg": 2.75,
"contribution": "2;2;2;2",
"contribution_avg": 2,
"presentation": "2;2;3;3",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:01.081438"
} | {
"id": "AZhWVPw1LL",
"metareview": "This paper proposes a RAG-like approach, with multiple tasks. Reviewers generally commented that the work lacked significant novelty compared to the original RAG, while the experiments lacked rigor. There was a healthy discussion which was not sufficient for the reviewers to inc... | {
"decision": "Reject"
} |
9WYMDgxDac | 2410.08174v1 | Sample then Identify: A General Framework for Risk Control and Assessment in Multimodal Large Language Models | {
"content": "## Abstract\n\nAbstract Multimodal Large Language Models (MLLMs) exhibit promising advancements across various tasks, yet they still encounter significant trustworthiness issues.\nPrior studies apply Split Conformal Prediction (SCP) in language modeling to construct prediction sets with statistical guar... | [
{
"id": "vIL1R9mDA5",
"initial_rating": 8,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "The paper introduces a two step risk control based framework extending split conformal prediction method for open ended and multimodal (videoQA) tasks. The method... | {
"rating": "5;5;6;8",
"rating_avg": 6,
"confidence": "4;3;4;4",
"confidence_avg": 3.75,
"soundness": "3;3;3;3",
"soundness_avg": 3,
"contribution": "2;3;3;3",
"contribution_avg": 2.75,
"presentation": "3;3;2;3",
"presentation_avg": 2.75
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Spotlight",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.082197"
} | {
"id": "Vw5BXoRw5B",
"metareview": "This paper presents a novel pipeline for risk assessment for open-ended LLM generated responses to visual question answering systems. This work addresses limitations of the current state of the art method of using Split Conformal Prediction (SCP) to construct estimates of the e... | {
"decision": "Accept (Spotlight)"
} |
9WbNpRuFuS | 2410.01103v1 | Approximately Aligned Decoding | {
"content": "## Abstract\n\nAbstract It is common to reject undesired outputs of Large Language Models (LLMs); however, current methods to do so require an excessive amount of computation, or severely distort the distribution of outputs.\nWe present a method to balance the distortion of the output distribution with ... | [
{
"id": "pEG3QZv8yO",
"initial_rating": 5,
"confidence": 3,
"soundness": 3,
"contribution": 2,
"presentation": 2,
"summary": "The paper proposes a method to speed up the recently proposed ASAp algorithm for constrained decoding by leveraging a connection to speculative decoding. This con... | {
"rating": "3;5;5;8",
"rating_avg": 5.25,
"confidence": "4;4;3;3",
"confidence_avg": 3.5,
"soundness": "3;4;3;3",
"soundness_avg": 3.25,
"contribution": "2;3;2;3",
"contribution_avg": 2.5,
"presentation": "1;2;2;4",
"presentation_avg": 2.25
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:01.082887"
} | {
"id": "lvgaIh5X1Z",
"metareview": "The paper proposes a novel method for constrained decoding in LLMs that balances output distortion and computational efficiency. The paper is well-written and presents ideas clearly, making it approachable for readers not familiar with the area. It also tackles an important pr... | {
"decision": "Reject"
} |
9Wghi9fKFA | 2410.08228v1 | Multi-Atlas Brain Network Classification through Consistency Distillation and Complementary Information Fusion | {
"content": "## Abstract\n\nAbstract In the realm of neuroscience, identifying distinctive patterns associated with neurological disorders via brain networks is crucial. Resting-state functional magnetic resonance imaging (fMRI) serves as a primary tool for mapping these networks by correlating blood-oxygen-level-de... | [
{
"id": "jUrVToAnTw",
"initial_rating": 8,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "This paper aims to classify brain networks from fMRI BOLD ROIs using multiple (potentially conflicting) atlases. The authors resolve conflicts with a deep neural ... | {
"rating": "3;5;5;8",
"rating_avg": 5.25,
"confidence": "3;5;4;3",
"confidence_avg": 3.75,
"soundness": "2;2;3;3",
"soundness_avg": 2.5,
"contribution": "2;2;2;3",
"contribution_avg": 2.25,
"presentation": "3;3;3;3",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:01.083615"
} | {
"id": "daXJEGqY1o",
"metareview": "This paper introduces a transformer for fMRI data (presumably only resting state correlation matrices). They present classification task accuracy on four datasets (ABIDE, ADNI, PPMI, and Matai).\n\nReviewer responses are mixed, though discussion has improved scores. There remain... | {
"decision": "Reject"
} |
9XETcRsufZ | 2410.19034v1 | Mixture of Parrots: Experts improve memorization more than reasoning | {
"content": "## Abstract\n\nAbstract The Mixture-of-Experts (MoE) architecture enables a significant increase in the total number of model parameters with minimal computational overhead.\nHowever, it is not clear what performance tradeoffs, if any, exist between MoEs and standard dense transformers.\nIn this paper,\... | [
{
"id": "SMfBmycunH",
"initial_rating": 6,
"confidence": 3,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "The paper explores the effectiveness of Mixture-of-Experts (MoE) architectures compared to standard dense Transformers, focusing on their abilities in memorizatio... | {
"rating": "5;6;8;8",
"rating_avg": 6.75,
"confidence": "4;3;4;4",
"confidence_avg": 3.75,
"soundness": "2;3;4;4",
"soundness_avg": 3.25,
"contribution": "2;2;4;3",
"contribution_avg": 2.75,
"presentation": "3;3;4;4",
"presentation_avg": 3.5
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.084603"
} | {
"id": "J5b29WPulW",
"metareview": "(a) Summary of scientific Claims and Findings:\nThe paper investigates performance tradeoffs between Mixture-of-Experts (MoE) and dense transformer architectures. The key findings are:\n- As the number of experts increases (while fixing active parameters), memorization/recall pe... | {
"decision": "Accept (Poster)"
} |
9XabBgqFgy | 2410.06373v1 | Unveiling the Backbone-Optimizer Coupling Bias in Visual Representation Learning | {
"content": "## Abstract\n\nAbstract This paper delves into the interplay between vision backbones and optimizers, unvealing an inter-dependent phenomenon termed b ackbone- o ptimizer c oupling b ias (BOCB).\nWe observe that canonical CNNs, such as VGG and ResNet, exhibit a marked co-dependency with SGD families, wh... | [
{
"id": "jJLvn9icYJ",
"initial_rating": 5,
"confidence": 3,
"soundness": 2,
"contribution": 2,
"presentation": 2,
"summary": "The paper studies the interaction between popular optimizers and vision backbone architectures. The experiments are conducted over the product of 16 architectures... | {
"rating": "5;5;5",
"rating_avg": 5,
"confidence": "3;3;3",
"confidence_avg": 3,
"soundness": "2;3;2",
"soundness_avg": 2.3333333333333335,
"contribution": "2;2;2",
"contribution_avg": 2,
"presentation": "3;2;2",
"presentation_avg": 2.3333333333333335
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:01.085539"
} | {
"id": "j926BUbLJO",
"metareview": "This paper investigates the interaction between vision backbone architectures and various optimizers, introducing the concept of Backbone-Optimizer Coupling Bias to explain how certain combinations of optimizers and architectures perform better than others. Through a comprehensi... | {
"decision": "Reject"
} |
9XprjIqkBI | 2405.18741v2 | Genshin: General Shield for Natural Language Processing with Large Language Models | {
"content": "## Abstract\n\nAbstract Large language models (LLMs) like ChatGPT, Gemini, or LLaMA have been trending recently, demonstrating considerable advancement and generalizability power in countless domains.\nHowever, LLMs create an even bigger black box exacerbating opacity, with interpretability limited to f... | [
{
"id": "4WFBPvJsK4",
"initial_rating": 1,
"confidence": 5,
"soundness": 1,
"contribution": 1,
"presentation": 1,
"summary": "This paper proposes a General Shield for Natural Language Processing with Large Language Models (Genshin), utilizing LLMs as defensive one-time plug-ins. It inclu... | {
"rating": "1;3;3;3;5",
"rating_avg": 3,
"confidence": "5;4;4;4;3",
"confidence_avg": 4,
"soundness": "1;3;1;2;2",
"soundness_avg": 1.8,
"contribution": "1;2;1;2;2",
"contribution_avg": 1.6,
"presentation": "1;3;2;1;2",
"presentation_avg": 1.8
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:01.086452"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
9aTZf71uiD | 2405.17680v1 | Sports-Traj: A Unified Trajectory Generation Model for Multi-Agent Movement in Sports | {
"content": "## Abstract\n\nAbstract Understanding multi-agent behavior is critical across various fields. The conventional approach involves analyzing agent movements through three primary tasks: trajectory prediction, imputation, and spatial-temporal recovery. Considering the unique input formulation and constrain... | [
{
"id": "UkKHbqWE0T",
"initial_rating": 5,
"confidence": 4,
"soundness": 2,
"contribution": 2,
"presentation": 3,
"summary": "The authors focus on the trajectory prediction in sport scenes. They propose a trajectory generation model. They extend the Mamba model into a bidirectional tempo... | {
"rating": "5;6;6;6",
"rating_avg": 5.75,
"confidence": "4;3;3;4",
"confidence_avg": 3.5,
"soundness": "2;2;2;3",
"soundness_avg": 2.25,
"contribution": "2;3;3;3",
"contribution_avg": 2.75,
"presentation": "3;3;4;3",
"presentation_avg": 3.25
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.087323"
} | {
"id": "javihAZZvj",
"metareview": "The authors designed a novel approach for handling multiple tasks, which include trajectory prediction, imputation, and spatial-temporal recovery, for multi-agent motion analysis. All the four reviewers pointed out that the method is good and well designed, and thus all recommen... | {
"decision": "Accept (Poster)"
} |
9bMZ29SPVx | 2410.11215v1 | A CLIP-Powered Framework for Robust and Generalizable Data Selection | {
"content": "## Abstract\n\nAbstract Large-scale datasets have been pivotal to the advancements of deep learning models in recent years, but training on such large datasets invariably incurs substantial storage and computational overhead.\nMeanwhile, real-world datasets often contain redundant and noisy data, imposi... | [
{
"id": "zB0aj0QAlU",
"initial_rating": 6,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "This paper introduces a CLIP-based multimodal data selection framework that enhances the robustness and generalizability of data selection by leveraging both imag... | {
"rating": "6;6;6;8",
"rating_avg": 6.5,
"confidence": "3;3;4;3",
"confidence_avg": 3.25,
"soundness": "3;3;3;4",
"soundness_avg": 3.25,
"contribution": "3;3;3;4",
"contribution_avg": 3.25,
"presentation": "3;3;3;4",
"presentation_avg": 3.25
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Spotlight",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.088034"
} | {
"id": "NTuLD1iG4a",
"metareview": "This work proposes a new framework for data selection to address the computational overhead and the impact of noisy data when training deep learning models. The key idea is to leverage multimodal information to better select important data and this is achieved by utilising the C... | {
"decision": "Accept (Spotlight)"
} |
9c96mGtQVR | 2405.17049v1 | Verifying Properties of Binary Neural Networks Using Sparse Polynomial Optimization | {
"content": "## Abstract\n\nAbstract This paper explores methods for verifying the properties of Binary Neural Networks (BNNs), focusing on robustness against adversarial attacks. Despite their lower computational and memory needs, BNNs, like their full-precision counterparts, are also sensitive to input perturbatio... | [
{
"id": "SEy5cRVHey",
"initial_rating": 6,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "The paper presents a novel approach for verifying properties of Binary Neural Networks (BNNs), particularly in the context of robustness against adversarial attac... | {
"rating": "3;3;6;6",
"rating_avg": 4.5,
"confidence": "3;4;3;3",
"confidence_avg": 3.25,
"soundness": "2;3;3;3",
"soundness_avg": 2.75,
"contribution": "3;2;3;3",
"contribution_avg": 2.75,
"presentation": "1;3;3;3",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.088631"
} | {
"id": "70gzqaF0Jm",
"metareview": "The paper proposes a new approach for verifying binary neural networks. It is based on a novel formulation of semidefinite programming (SDP), derived from sparse polynomial optimization. The angle to develop this new formulation is novel, and the paper also includes good insight... | {
"decision": "Accept (Poster)"
} |
9ccZzuix2D | 2403.07854v2 | Distilling the Knowledge in Data Pruning | {
"content": "## Abstract\n\nAbstract With the increasing size of datasets used for training neural networks, data pruning becomes an attractive field of research.\nHowever, most current data pruning algorithms are limited in their ability to preserve accuracy compared to models trained on the full data, especially i... | [
{
"id": "gEDCjmxgRv",
"initial_rating": 3,
"confidence": 3,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "This paper presents an in-depth investigation into the use of knowledge distillation (KD) for training models on pruned datasets. It provides a comprehensive anal... | {
"rating": "3;5;5",
"rating_avg": 4.333333333333333,
"confidence": "3;4;4",
"confidence_avg": 3.6666666666666665,
"soundness": "3;2;2",
"soundness_avg": 2.3333333333333335,
"contribution": "2;2;2",
"contribution_avg": 2,
"presentation": "3;4;2",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:01.089242"
} | {
"id": "f65sCz3Vq1",
"metareview": "This paper investigates the use of knowledge distillation (KD) to improve the training of neural networks on pruned datasets. It demonstrates that incorporating KD from a teacher model trained on the full dataset can enhance model performance across various datasets, pruning met... | {
"decision": "Reject"
} |
9fMNxWDZsP | 2408.13438v2 | Explainable Concept Generation through Vision-Language Preference Learning | {
"content": "## Abstract\n\nAbstract Concept-based explanations have become a popular choice for explaining deep neural networks post-hoc because, unlike most other explainable AI techniques, they can be used to test high-level visual “concepts” that are not directly related to feature attributes. For instance, the ... | [
{
"id": "91hQqNvqC9",
"initial_rating": 5,
"confidence": 3,
"soundness": 2,
"contribution": 2,
"presentation": 2,
"summary": "This paper focuses on the generation of concept images to explain black-box image classification models. It proposes a reinforcement learning-based preference opt... | {
"rating": "3;3;5;5;8;8",
"rating_avg": 5.333333333333333,
"confidence": "3;2;3;3;4;3",
"confidence_avg": 3,
"soundness": "1;3;3;2;3;3",
"soundness_avg": 2.5,
"contribution": "1;2;3;2;3;3",
"contribution_avg": 2.3333333333333335,
"presentation": "2;3;2;2;3;3",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:01.090607"
} | {
"id": "SOYOn9PLVr",
"metareview": "Scientific Claims and Findings:\n\nThis paper introduces a novel algorithm called Reinforcement Learning-based Preference Optimization (RLPO) to automatically generate visual concepts that explain the decisions of deep neural networks (DNNs). The algorithm addresses the limitat... | {
"decision": "Reject"
} |
9h45qxXEx0 | 2410.01209v1 | Debiasing Federated Learning with Correlated Client Participation | {
"content": "## Abstract\n\nAbstract In cross-device federated learning (FL) with millions of mobile clients, only a small subset of clients participate in training in every communication round, and Federated Averaging (FedAvg) is the most popular algorithm in practice. Existing analyses of FedAvg usually assume the... | [
{
"id": "rb4WNdTvaP",
"initial_rating": 8,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 2,
"summary": "This paper studied the federated learning problem with partial client participation. In particular, it focused on the case where there is minimum separate between... | {
"rating": "3;5;6;8",
"rating_avg": 5.5,
"confidence": "3;3;3;4",
"confidence_avg": 3.25,
"soundness": "2;3;3;3",
"soundness_avg": 2.75,
"contribution": "2;3;3;3",
"contribution_avg": 2.75,
"presentation": "2;2;3;2",
"presentation_avg": 2.25
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.091538"
} | {
"id": "4HQXrgqZNo",
"metareview": "This introduces a theoretical framework modeling client participation in federated learning (FL) as a Markov chain, addressing non-uniform and correlated participation across rounds. The authors analyze scenarios where clients must wait a minimum number of rounds before re-parti... | {
"decision": "Accept (Poster)"
} |
9h5paerJxC | 2408.05916v1 | Cluster-Segregate-Perturb (CSP): A Model-agnostic Explainability Pipeline for Spatiotemporal Land Surface Forecasting Models | {
"content": "## Abstract\n\nAbstract Satellite images have become increasingly valuable for modelling regional climate change effects. Earth surface forecasting represents one such task that integrates satellite images with meteorological data to capture the joint evolution of regional climate change effects. Howeve... | [
{
"id": "b77gnyiWbs",
"initial_rating": 3,
"confidence": 4,
"soundness": 3,
"contribution": 1,
"presentation": 2,
"summary": "This paper introduces the Cluster-Segregate-Perturb (CSP) pipeline as an approach for achieving explainability in land surface forecasting. The CSP pipeline inclu... | {
"rating": "3;3;3;5",
"rating_avg": 3.5,
"confidence": "4;4;4;4",
"confidence_avg": 4,
"soundness": "1;2;3;3",
"soundness_avg": 2.25,
"contribution": "2;1;1;2",
"contribution_avg": 1.5,
"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": "2026-01-14T22:16:01.092368"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
9hpcTgztk8 | 2310.11085v4 | Document-Level In-Context Few-Shot Relation Extraction via Pre-Trained Language Models | {
"content": "## Abstract\n\nAbstract Document-level relation extraction aims at inferring structured human knowledge from textual documents. State-of-the-art methods for this task use pre-trained language models (LMs) via fine-tuning, yet fine-tuning is computationally expensive and cannot adapt to new relation type... | [
{
"id": "UH8MvITpg3",
"initial_rating": 3,
"confidence": 4,
"soundness": 3,
"contribution": 2,
"presentation": 2,
"summary": "This paper presents a new framework called REPLM for document-level few-shot relation extraction using pre-trained language models (LMs). The key idea is to refor... | {
"rating": "3;3;5;6",
"rating_avg": 4.25,
"confidence": "4;4;4;3",
"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;2;3",
"presentation_avg": 2.25
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:01.093262"
} | {
"id": "2tJLhXShH0",
"metareview": "The paper presents REPLM, a novel framework for document-level in-context few-shot relation extraction using pre-trained language models (LMs) without fine-tuning. It aims to address limitations of existing methods by eliminating the need for named entity recognition and human-a... | {
"decision": "Reject"
} |
9htTvHkUhh | 2410.11933v1 | Beyond Sequence: Impact of Geometric Context for RNA Property Prediction | {
"content": "## Abstract\n\nAbstract Accurate prediction of RNA properties, such as stability and interactions, is crucial for advancing our understanding of biological processes and developing RNA-based therapeutics. RNA structures can be represented as 1D sequences, 2D topological graphs, or 3D all-atom models, ea... | [
{
"id": "vPD9EMICBV",
"initial_rating": 6,
"confidence": 4,
"soundness": 2,
"contribution": 2,
"presentation": 3,
"summary": "This paper introduce a newly curated set of RNA datasets with enhanced 2D and 3D structural annotations, providing a resource for model evaluation on RNA data. Th... | {
"rating": "3;3;5",
"rating_avg": 3.6666666666666665,
"confidence": "4;4;4",
"confidence_avg": 4,
"soundness": "2;2;2",
"soundness_avg": 2,
"contribution": "2;2;2",
"contribution_avg": 2,
"presentation": "2;3;3",
"presentation_avg": 2.6666666666666665
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.094474"
} | {
"id": "7DTMvXPZxD",
"metareview": "This paper introduces a curated set of RNA datasets with enhanced 2D and 3D structural annotations, providing a valuable resource for evaluating RNA property prediction models. It systematically investigates the impact of incorporating geometric information, showing that models ... | {
"decision": "Accept (Poster)"
} |
9iN8p1Xwtg | 2409.17422v1 | Discovering the Gems in Early Layers: Accelerating Long-Context LLMs with 1000x Input Token Reduction | {
"content": "## Abstract\n\nLarge Language Models (LLMs) have demonstrated remarkable capabilities in handling long context inputs, but this comes at the cost of increased computational resources and latency. Our research introduces a novel approach for the long context bottleneck to accelerate LLM inference and red... | [
{
"id": "fKIJ2W5z7J",
"initial_rating": 5,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "This paper introduces GemFilter, which leverages early layers of large language models (LLMs) to compress input tokens and accelerate inference. GemFilter identif... | {
"rating": "3;5;5;6",
"rating_avg": 4.75,
"confidence": "4;4;4;3",
"confidence_avg": 3.75,
"soundness": "2;2;3;3",
"soundness_avg": 2.5,
"contribution": "2;2;3;3",
"contribution_avg": 2.5,
"presentation": "2;3;3;3",
"presentation_avg": 2.75
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:01.095132"
} | {
"id": "M3FimuTXyE",
"metareview": "The paper proposes a novel approach, GemFilter, designed to accelerate inference and reduce memory consumption in large language models (LLMs) by utilizing early layers of the model to identify and compress relevant input tokens. This method aims to address the long-context bott... | {
"decision": "Reject"
} |
9juyeCqL0u | 2310.15117v1 | Causal Order: The Key to Leveraging Imperfect Experts in Causal Inference | {
"content": "## Abstract\n\nAbstract At the core of causal inference lies the challenge of determining reliable causal graphs solely based on observational data. Since the well-known backdoor criterion depends on the graph, any errors in the graph can propagate downstream to effect inference. In this work, we initia... | [
{
"id": "9yqwr13niE",
"initial_rating": 6,
"confidence": 4,
"soundness": 4,
"contribution": 3,
"presentation": 3,
"summary": "This paper considers optimal ways of querying imperfect experts (e.g., LLMs) when the aim is to discover a causal DAG over a set of variables. The key proposed id... | {
"rating": "5;5;6;6;6",
"rating_avg": 5.6,
"confidence": "3;3;3;4;4",
"confidence_avg": 3.4,
"soundness": "2;2;3;3;4",
"soundness_avg": 2.8,
"contribution": "1;2;3;3;3",
"contribution_avg": 2.4,
"presentation": "3;2;4;2;3",
"presentation_avg": 2.8
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.095768"
} | {
"id": "8gg2VtyYzV",
"metareview": "The paper introduces a strategy for querying imperfect experts (such as LLMs) for causal discovery, focusing on causal ordering over graphical relationships.\n\n\nStrengths:\n\n+ The paper proposed triplet querying strategy for causal discovery.\n\n\n+ The inclusion of both huma... | {
"decision": "Accept (Poster)"
} |
9klRFLY2TT | 2402.08777v3 | DNABERT-S: Pioneering Species Differentiation with Species-Aware DNA Embeddings | {
"content": "## Abstract\n\nAbstract We introduce DNABERT-S, a tailored genome model that develops species-aware embeddings to naturally cluster and segregate DNA sequences of different species in the embedding space.\nDifferentiating species from genomic sequences (i.e., DNA and RNA) is vital yet challenging, since... | [
{
"id": "bXAuffr2Zy",
"initial_rating": 5,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "The paper introduces DNABERT-S, a genome model focused on species-aware DNA embeddings to differentiate and cluster DNA sequences by species effectively. Building... | {
"rating": "5;6;6",
"rating_avg": 5.666666666666667,
"confidence": "4;4;3",
"confidence_avg": 3.6666666666666665,
"soundness": "3;3;2",
"soundness_avg": 2.6666666666666665,
"contribution": "3;3;3",
"contribution_avg": 3,
"presentation": "3;3;3",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:01.096472"
} | {
"id": "7I3KkveCqM",
"metareview": "The authors introduce techniques like DNA-Dropout and DNA-Double to improve embedding representations, enhancing the model's ability to distinguish between species. The paper utilizes contrastive learning techniques, such as MI-Mix and C2LR, to gradually introduce increasingly d... | {
"decision": "Reject"
} |
9ljHiYuRHl | 2410.23884v1 | Failure Modes of LLMs for Causal Reasoning on Narratives | {
"content": "## Abstract\n\nAbstract In this work, we investigate the causal reasoning abilities of large language\nmodels (LLMs) through the representative problem of inferring causal\nrelationships from narratives.\nWe find that even state-of-the-art language models rely\non unreliable shortcuts,\nboth in terms of... | [
{
"id": "Mgiv2q3sQi",
"initial_rating": 6,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "The paper investigates the causal reasoning capabilities of recent LLMs; through various settings; via synthetic, semi-synthetic and real world narratives, compar... | {
"rating": "3;3;5;6",
"rating_avg": 4.25,
"confidence": "3;4;3;4",
"confidence_avg": 3.5,
"soundness": "2;2;2;3",
"soundness_avg": 2.25,
"contribution": "2;1;2;3",
"contribution_avg": 2,
"presentation": "2;1;3;3",
"presentation_avg": 2.25
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:01.097241"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
9mBodivRIo | 2410.06437v1 | LocoVR: Multiuser Indoor Locomotion Dataset in Virtual Reality | {
"content": "## Abstract\n\nAbstract Understanding human locomotion is crucial for AI agents such as robots, particularly in complex indoor home environments. Modeling human trajectories in these spaces requires insight into how individuals maneuver around physical obstacles and manage social navigation dynamics. Th... | [
{
"id": "O0mTQPCCBd",
"initial_rating": 3,
"confidence": 4,
"soundness": 2,
"contribution": 2,
"presentation": 3,
"summary": "This paper introduces a dataset for multi-user indoor navigation collected in a Virtual Reality (VR) environment. The dataset includes 7,071 trajectories with 2.5... | {
"rating": "3;5;6;6;8",
"rating_avg": 5.6,
"confidence": "4;3;4;3;4",
"confidence_avg": 3.6,
"soundness": "2;2;3;3;4",
"soundness_avg": 2.8,
"contribution": "2;2;3;3;4",
"contribution_avg": 2.8,
"presentation": "3;3;3;3;4",
"presentation_avg": 3.2
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.098127"
} | {
"id": "zxbeov6kW4",
"metareview": "The submission is about a new dataset of two-person trajectories captured in VR. Reviewers acknowledged the usefulness of the dataset; they also raised some concerns, primarily about the validity of data collected in VR. Post rebuttal, most reviewers were convinced and support... | {
"decision": "Accept (Poster)"
} |
9mjZ800m7Y | 2402.18213v2 | Multi-objective Differentiable Neural Architecture Search | {
"content": "## Abstract\n\nAbstract Pareto front profiling in multi-objective optimization (MOO), i.e. finding a diverse set of Pareto optimal solutions, is challenging, especially with expensive objectives like neural network training. Typically, in MOO neural architecture search (NAS), we aim to balance performan... | [
{
"id": "TdaDVyq7SI",
"initial_rating": 5,
"confidence": 4,
"soundness": 2,
"contribution": 2,
"presentation": 2,
"summary": "The paper (MODNAS) presents an approach to Neural Architecture Search (NAS) that balances competing objectives—like performance, latency, and energy efficiency—ac... | {
"rating": "5;5;6;8",
"rating_avg": 6,
"confidence": "4;4;4;3",
"confidence_avg": 3.75,
"soundness": "3;2;3;4",
"soundness_avg": 3,
"contribution": "2;2;2;4",
"contribution_avg": 2.5,
"presentation": "2;2;3;4",
"presentation_avg": 2.75
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.099071"
} | {
"id": "rAoFIe6xHd",
"metareview": "This paper presents a hypernetwork-based method for hardware-aware neural architecture search and demonstrates zero-shot generalization to new devices. The strengths of this paper include the detailed experiments that tested the proposed MODNAS on 19 hardware devices and showed ... | {
"decision": "Accept (Poster)"
} |
9oMB6wnFYM | 2401.14404v1 | Deconstructing Denoising Diffusion Models for Self-Supervised Learning | {
"content": "## Abstract\n\nAbstract In this study, we examine the representation learning abilities of Denoising Diffusion Models (DDM) that were originally purposed for image generation. Our philosophy is to deconstruct a DDM, gradually transforming it into a classical Denoising Autoencoder (DAE). This deconstruct... | [
{
"id": "wKnaFF4nV3",
"initial_rating": 6,
"confidence": 2,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "In this paper, the author studied the representation learning abilities of denoising-autoencoder-based diffusion models (DDMs). Throughout extensive ablation stud... | {
"rating": "6;6;6;6",
"rating_avg": 6,
"confidence": "3;4;4;2",
"confidence_avg": 3.25,
"soundness": "3;3;3;3",
"soundness_avg": 3,
"contribution": "2;2;3;3",
"contribution_avg": 2.5,
"presentation": "3;2;3;3",
"presentation_avg": 2.75
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.099870"
} | {
"id": "pFjOha7OzW",
"metareview": "In this paper the authors conduct a sequence of experiments which progressively transforms a Denoising Diffusion Model in a Denoising Autoencoder to understand how the various aspects of the model impact the overall performance and to identify which components are essential (or ... | {
"decision": "Accept (Poster)"
} |
9oq0iY2Jxx | 2405.17618v2 | Symmetric Reinforcement Learning Loss for Robust Learning on Diverse Tasks and Model Scales | {
"content": "## Abstract\n\nAbstract Reinforcement learning (RL) training is inherently unstable due to factors such as moving targets and high gradient variance. Reinforcement Learning from Human Feedback (RLHF) and Reinforcement Learning from AI Feedback (RLAIF) can introduce additional difficulty. Differing prefe... | [
{
"id": "PKgYRxR4wt",
"initial_rating": 3,
"confidence": 4,
"soundness": 3,
"contribution": 2,
"presentation": 2,
"summary": "This paper considers the problem of a noisy signal affecting the policy updates when using actor critic algorithms for policy improvement. It takes inspiration fr... | {
"rating": "3;3;3;5",
"rating_avg": 3.5,
"confidence": "4;3;4;2",
"confidence_avg": 3.25,
"soundness": "1;3;3;2",
"soundness_avg": 2.25,
"contribution": "1;2;2;2",
"contribution_avg": 1.75,
"presentation": "2;1;2;1",
"presentation_avg": 1.5
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:01.100500"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
9qpdDiDQ2H | 2410.03074v1 | MetaOOD: Automatic Selection of OOD Detection Models | {
"content": "## Abstract\n\nAbstract How can we automatically select an out-of-distribution (OOD) detection model for various underlying tasks? This is crucial for maintaining the reliability of open-world applications by identifying data distribution shifts, particularly in critical domains such as online transacti... | [
{
"id": "3eCYPvM311",
"initial_rating": 6,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "In this paper, the authors propose MetaOOD, which utilizes meta-learning to select an OOD detection model automatically. The motivation is that each OOD detection... | {
"rating": "3;6;6;6",
"rating_avg": 5.25,
"confidence": "4;5;5;4",
"confidence_avg": 4.5,
"soundness": "1;3;2;1",
"soundness_avg": 1.75,
"contribution": "1;3;3;3",
"contribution_avg": 2.5,
"presentation": "1;2;2;3",
"presentation_avg": 2
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.101239"
} | {
"id": "baJWU1DD33",
"metareview": "This paper presents a meta-learning framework for model selection of out-of-distribution detection models. The core idea is to use embeddings from language models to represent datasets and models as described by textual descriptions. The framework is evaluated on image-based OOD... | {
"decision": "Accept (Poster)"
} |
9rtlfjWMXI | 2408.09181v1 | PADetBench: Towards Benchmarking Physical Attacks against Object Detection | {
"content": "## Abstract\n\nAbstract Physical attacks against object detection have gained increasing attention due to their significant practical implications.\nHowever, conducting physical experiments is extremely time-consuming and labor-intensive.\nMoreover, physical dynamics and cross-domain transformation are ... | [
{
"id": "WGAUN3h9Xn",
"initial_rating": 5,
"confidence": 5,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "This paper provides an end-to-end pipeline to evaluate physical adversarial examples with different parameters, including environments, vehicle and pedestrian mod... | {
"rating": "3;5;5;6",
"rating_avg": 4.75,
"confidence": "5;3;5;4",
"confidence_avg": 4.25,
"soundness": "2;3;3;3",
"soundness_avg": 2.75,
"contribution": "2;3;3;3",
"contribution_avg": 2.75,
"presentation": "1;1;3;3",
"presentation_avg": 2
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:01.101974"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
9soA8GWQ9g | 2411.00666v1 | Beyond the Boundaries of Proximal Policy Optimization | {
"content": "## Abstract\n\nAbstract Proximal policy optimization (PPO) is a widely-used algorithm for on-policy reinforcement learning. This work offers an alternative perspective of PPO, in which it is decomposed into the inner-loop estimation of update vectors, and the outer-loop application of updates using grad... | [
{
"id": "8tiru1Q3Sf",
"initial_rating": 3,
"confidence": 2,
"soundness": 3,
"contribution": 1,
"presentation": 3,
"summary": "The paper designs a novel framework, called outer-PPO, to further modify PPO’s trust region gradients through an outer loop. PPO conducts several gradient updates... | {
"rating": "3;3;5;6",
"rating_avg": 4.25,
"confidence": "5;2;2;4",
"confidence_avg": 3.25,
"soundness": "2;3;2;3",
"soundness_avg": 2.5,
"contribution": "1;1;2;2",
"contribution_avg": 1.5,
"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": "2026-01-14T22:16:01.102647"
} | {
"id": "9ZiIcJAtkl",
"metareview": "The paper proposes a new algorithm that creates an inner and outer loop PPO method where the inner loop estimates an update direction and the outer loop actually changes the policy. The paper performs a non-insignificant amount of hyperparameter tuning to make a performance comp... | {
"decision": "Reject"
} |
A0VvDN4arV | 2406.00059v2 | Conveyor: Efficient Tool-aware LLM Serving with Tool Partial Execution | {
"content": "## Abstract\n\nAbstract The complexity of large language model (LLM) serving workloads has substantially increased due to the integration with external tool invocations, such as ChatGPT plugins.\nIn this paper, we identify a new opportunity for efficient LLM serving for requests that trigger tools: tool... | [
{
"id": "tSh7Emo6hL",
"initial_rating": 3,
"confidence": 3,
"soundness": 1,
"contribution": 1,
"presentation": 2,
"summary": "The paper introduces Conveyor, an efficient LLM serving system optimized for the latency of workloads involving tool executions. Conveyor achieves this by separat... | {
"rating": "3;3;3;5",
"rating_avg": 3.5,
"confidence": "4;4;3;4",
"confidence_avg": 3.75,
"soundness": "2;2;1;2",
"soundness_avg": 1.75,
"contribution": "2;2;1;2",
"contribution_avg": 1.75,
"presentation": "3;3;2;3",
"presentation_avg": 2.75
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:01.103771"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
A0W7VCSQev | 2410.02343v1 | Listening to the Wise Few: Select-and-Copy Attention Heads for Multiple-Choice QA | {
"content": "## Abstract\n\nAbstract A standard way to evaluate the abilities of LLM involves presenting a multiple-choice question and selecting the option with the highest logit as the model’s predicted answer. However, such a format for evaluating LLMs has limitations, since even if the model knows the correct an... | [
{
"id": "mIehyoALPB",
"initial_rating": 3,
"confidence": 5,
"soundness": 2,
"contribution": 3,
"presentation": 4,
"summary": "The widely used evaluation for large language models, multiple-choice question answering (MCQA), is very brittle, especially for small models -- existing works sh... | {
"rating": "3;5;5;6",
"rating_avg": 4.75,
"confidence": "5;4;5;3",
"confidence_avg": 4.25,
"soundness": "2;3;4;3",
"soundness_avg": 3,
"contribution": "3;3;2;3",
"contribution_avg": 2.75,
"presentation": "4;3;4;2",
"presentation_avg": 3.25
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:01.104430"
} | {
"id": "Hj9UyKKrch",
"metareview": "This study introduces a novel method aimed at enhancing the evaluation of large language models (LLMs) in multiple-choice question answering (MCQA) by leveraging select-and-copy heads, which are specific attention heads. These heads consistently extract pertinent information, th... | {
"decision": "Reject"
} |
A2rfALKFBg | 2410.00340v3 | Sparse Attention Decomposition Applied to Circuit Tracing | {
"content": "## Abstract\n\nAbstract Many papers have shown that attention heads work in conjunction with each other to perform complex tasks. It’s frequently assumed that communication between attention heads is via the addition of specific features to token residuals.\nIn this work we seek to isolate and identify ... | [
{
"id": "WG4xdsumGb",
"initial_rating": 5,
"confidence": 4,
"soundness": 3,
"contribution": 2,
"presentation": 2,
"summary": "The authors explored some of the technical details of GPT-2 through SPARSE ATTENTION DECOMPOSITION. Their tracing study reveals considerable detail not present in... | {
"rating": "3;5;6",
"rating_avg": 4.666666666666667,
"confidence": "4;4;3",
"confidence_avg": 3.6666666666666665,
"soundness": "2;3;3",
"soundness_avg": 2.6666666666666665,
"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": "2026-01-14T22:16:01.105155"
} | {
"id": "qvYWJx6Nck",
"metareview": "The paper presents an approach for analyzing communication between attention heads in transformer models using SVD, but reviewers raised several critical concerns. Reviewers noted the limitation of the analysis, which is confined to GPT-2 and does not consider other mainstream m... | {
"decision": "Reject"
} |
A4aG3XeIO7 | 2410.05140v2 | Tuning-Free Bilevel Optimization: New Algorithms and Convergence Analysis | {
"content": "## Abstract\n\nAbstract Bilevel optimization has recently attracted considerable attention due to its abundant applications in machine learning problems. However, existing methods rely on prior knowledge of problem parameters to determine stepsizes, resulting in significant effort in tuning stepsizes wh... | [
{
"id": "uH067qQDGn",
"initial_rating": 6,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "This paper has proposed two tuning-free algorithms for stochastic bilevel optimization, D-TFBO and S-TFBO, which eliminates the need for stepsize tuning that depe... | {
"rating": "5;6;6;6",
"rating_avg": 5.75,
"confidence": "2;3;3;4",
"confidence_avg": 3,
"soundness": "3;3;3;3",
"soundness_avg": 3,
"contribution": "2;3;3;3",
"contribution_avg": 2.75,
"presentation": "3;3;3;3",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.106301"
} | {
"id": "xpABMtysC3",
"metareview": "This paper studies a practically important problem on bilevel optimization with adaptive tuning of stepsizes, which will reduce significant efforts in tuning stepsizes when some of the problem parameters are unknown. The paper proposes two novel tuning-free algorithms, where one... | {
"decision": "Accept (Poster)"
} |
A4eCzSohhx | 2406.05753v4 | Grounding Continuous Representations in Geometry: Equivariant Neural Fields | {
"content": "## Abstract\n\nAbstract Conditional Neural Fields (CNFs) are increasingly being leveraged as continuous signal representations, by associating each data-sample with a latent variable that conditions a shared backbone Neural Field (NeF) to reconstruct the sample. However, existing CNF architectures face ... | [
{
"id": "uNjxp1uq5w",
"initial_rating": 3,
"confidence": 4,
"soundness": 2,
"contribution": 2,
"presentation": 2,
"summary": "This paper proposes equivariant conditional neural fields based on steerable networks. Architecture-wise, this paper proposes equivariant cross-attention layers ... | {
"rating": "3;6;6;8",
"rating_avg": 5.75,
"confidence": "4;4;3;3",
"confidence_avg": 3.5,
"soundness": "2;2;3;4",
"soundness_avg": 2.75,
"contribution": "2;2;3;4",
"contribution_avg": 2.75,
"presentation": "2;3;3;4",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.107538"
} | {
"id": "ufQZpLiFyb",
"metareview": "This paper proposes a variant of conditional neural fields by utilizing a latent point cloud rather than a global latent for a neural field. The authors claimed locality and steerability for the resulting Equivariant Neural Fields and demonstrated the effectiveness of latent for... | {
"decision": "Accept (Poster)"
} |
A72sZWB66Q | 2410.06044v1 | HyperDet: Generalizable Detection of Synthesized Images by Generating and Merging A Mixture of Hyper LoRAs | {
"content": "## Abstract\n\nAbstract The emergence of diverse generative vision models has recently enabled the synthesis of visually realistic images, underscoring the critical need for effectively detecting these generated images from real photos. Despite advances in this field, existing detection approaches often... | [
{
"id": "YGhwo6VDPm",
"initial_rating": 1,
"confidence": 4,
"soundness": 1,
"contribution": 2,
"presentation": 3,
"summary": "The paper proposes HyperDet, a mixture of experts approach towards the detection of generated images. The basic idea of the introduced method is to apply a pre-pr... | {
"rating": "1;3;5;5;5",
"rating_avg": 3.8,
"confidence": "4;4;4;2;4",
"confidence_avg": 3.6,
"soundness": "1;2;3;3;2",
"soundness_avg": 2.2,
"contribution": "2;2;3;3;2",
"contribution_avg": 2.4,
"presentation": "3;2;3;2;3",
"presentation_avg": 2.6
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:01.108584"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
A7LTIuhH4k | 2410.02123v1 | Approximating Multiple Robust Optimization Solutions in One Pass via Proximal Point Methods | {
"content": "## Abstract\n\nAbstract Robust optimization provides a principled and unified framework to model many problems in modern operations research and computer science applications, such as risk measures minimization and adversarially robust machine learning. To use a robust solution (e.g., to implement an in... | [
{
"id": "YSkBeqvcMv",
"initial_rating": 5,
"confidence": 2,
"soundness": 2,
"contribution": 2,
"presentation": 2,
"summary": "This paper studies robust optimization with an uncertainty set, and proposes a more efficient way for computing Pareto efficient robust solutions based on proxima... | {
"rating": "5;5;5;5",
"rating_avg": 5,
"confidence": "4;2;2;2",
"confidence_avg": 2.5,
"soundness": "3;3;2;2",
"soundness_avg": 2.5,
"contribution": "1;2;2;2",
"contribution_avg": 1.75,
"presentation": "3;3;2;2",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:01.109240"
} | {
"id": "x4MKXVlXYQ",
"metareview": "A proximal point method based method is proposed to generate a series of Pareto efficient solutions with computational efficiency. However, the analysis is limited to a particular form of loss functions and is not obvious to the reviewers how to generalize to more general cases.... | {
"decision": "Reject"
} |
A9loYh0RgU | 2410.03794v1 | Repurposing Foundation Model for Generalizable Medical Time Series Classification | {
"content": "## Abstract\n\nAbstract Medical time series (MedTS) classification is critical for a wide range of healthcare applications such as Alzheimer’s Disease diagnosis. However, its real-world deployment is severely challenged by poor generalizability due to inter- and intra-dataset heterogeneity in MedTS, inc... | [
{
"id": "lKIu3kr7Tq",
"initial_rating": 3,
"confidence": 3,
"soundness": 3,
"contribution": 2,
"presentation": 2,
"summary": "The paper introduces FORMED, a new generalization method for foundation models. It introduces a novel mechnanism called re-purposing, designed for generalizable m... | {
"rating": "3;3;3;6",
"rating_avg": 3.75,
"confidence": "4;4;3;4",
"confidence_avg": 3.75,
"soundness": "2;2;3;3",
"soundness_avg": 2.5,
"contribution": "2;3;2;3",
"contribution_avg": 2.5,
"presentation": "2;2;2;4",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:01.110189"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
AAZ3vwyQ4X | 2410.22520v1 | Multimodal Structure Preservation Learning | {
"content": "## Abstract\n\nAbstract When selecting data to build machine learning models in practical applications, factors such as availability, acquisition cost, and discriminatory power are crucial considerations.\nDifferent data modalities often capture unique aspects of the underlying phenomenon, making their ... | [
{
"id": "dHg5HXsDjh",
"initial_rating": 3,
"confidence": 4,
"soundness": 2,
"contribution": 1,
"presentation": 2,
"summary": "This paper presents a multimodal framework called MSPL, which builds upon encoder decoder structure with extra regularizations form prediction task and structure ... | {
"rating": "1;3;3;3",
"rating_avg": 2.5,
"confidence": "4;3;5;4",
"confidence_avg": 4,
"soundness": "1;3;2;2",
"soundness_avg": 2,
"contribution": "1;2;2;1",
"contribution_avg": 1.5,
"presentation": "3;2;2;2",
"presentation_avg": 2.25
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:01.110934"
} | {
"id": "STvAsKqnQp",
"metareview": "This paper proposes a multimodal representation learning method called MSPL, which leverages an autoencoder with three loss functions: reconstruction, pretext task performance, and structural alignment. MSPL learns to preserve the structure of one modality, represented as a diss... | {
"decision": "Reject"
} |
ACSNlt77hq | 2410.05165v2 | Efficient Inference for Large Language Model-based Generative Recommendation | {
"content": "## Abstract\n\nAbstract Large Language Model (LLM)-based generative recommendation has achieved notable success, yet its practical deployment is costly particularly due to excessive inference latency caused by autoregressive decoding.\nFor lossless LLM decoding acceleration, Speculative Decoding (SD) ha... | [
{
"id": "EUpBByknPs",
"initial_rating": 6,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 4,
"summary": "This paper focuses on accelerating inference in LLM-based generative recommendation using speculative decoding. It highlights the challenges of applying speculati... | {
"rating": "3;5;6",
"rating_avg": 4.666666666666667,
"confidence": "4;4;4",
"confidence_avg": 4,
"soundness": "1;2;3",
"soundness_avg": 2,
"contribution": "3;2;3",
"contribution_avg": 2.6666666666666665,
"presentation": "2;2;4",
"presentation_avg": 2.6666666666666665
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.111600"
} | {
"id": "u5GFuMgGTt",
"metareview": "This paper studied the problem of speeding up using LLM-based model for top K recommendations, based on the framework of speculative decoding. \n\nStrength: \n1. An potentially useful technique for speeding up with LLM-based generative top K recommendation with some promising re... | {
"decision": "Accept (Poster)"
} |
AEglX9CHFN | 2411.01155v1 | HG-Adapter: Improving Pre-Trained Heterogeneous Graph Neural Networks with Dual Adapters | {
"content": "## Abstract\n\nAbstract The “pre-train, prompt-tuning” paradigm has demonstrated impressive performance for tuning pre-trained heterogeneous graph neural networks (HGNNs) by mitigating the gap between pre-trained models and downstream tasks. However, most prompt-tuning-based works may face at least two ... | [
{
"id": "GJGH209uY8",
"initial_rating": 6,
"confidence": 2,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "A dual-adapter approach has been introduced to graph prompt-tuning methods. The key insight of this work is to leverage dual adapters to capture both node feature... | {
"rating": "5;5;6",
"rating_avg": 5.333333333333333,
"confidence": "3;3;3",
"confidence_avg": 3,
"soundness": "2;2;3",
"soundness_avg": 2.3333333333333335,
"contribution": "2;2;3",
"contribution_avg": 2.3333333333333335,
"presentation": "3;3;2",
"presentation_avg": 2.6666666666666665
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.112472"
} | {
"id": "cNANAGICtK",
"metareview": "This paper claims that existing prompt-tuning-based works face two limitations: (i) the model may be insufficient to fit the graph structures well as they are generally ignored in the prompt-tuning stage, increasing the training error to decrease the generalization ability; and ... | {
"decision": "Accept (Poster)"
} |
AJM52ygi6Y | 2407.02020v3 | Decentralized Optimization with Coupled Constraints | {
"content": "## Abstract\n\nAbstract We consider the decentralized minimization of a separable objective ∑ i = 1 n f i ( x i ) superscript subscript 𝑖 1 𝑛 subscript 𝑓 𝑖 subscript 𝑥 𝑖 \\sum_{i=1}^{n}f_{i}(x_{i}) ∑ start_POSTSUBSCRIPT italic_i = 1 end_POSTSUBSCRIPT start_POSTSUPERSCRIPT italic_n end_POSTSUPERS... | [
{
"id": "lR3NOhLaPH",
"initial_rating": 6,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "This paper studies decentralized optimization with constraints. It establishes the lower bound for this problem, and develops algorithms to attain such lower boun... | {
"rating": "5;6;6;8",
"rating_avg": 6.25,
"confidence": "3;4;4;3",
"confidence_avg": 3.5,
"soundness": "3;3;3;4",
"soundness_avg": 3.25,
"contribution": "3;2;3;3",
"contribution_avg": 2.75,
"presentation": "2;3;3;3",
"presentation_avg": 2.75
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.113335"
} | {
"id": "oWExHo44Pe",
"metareview": "The paper addresses decentralized optimization of a separable objective with affine coupled constraints, relevant to resource allocation, systems control, and distributed machine learning. It establishes lower complexity bounds and proposes a first-order algorithm that achieves ... | {
"decision": "Accept (Poster)"
} |
AK1C55o4r7 | 2310.03940v5 | Beyond Random Augmentations: Pretraining with Hard Views | {
"content": "## Abstract\n\nAbstract Many Self-Supervised Learning (SSL) methods aim for model invariance to different image augmentations known as views . To achieve this invariance, conventional approaches make use of random sampling operations within the image augmentation pipeline. We hypothesize that the effica... | [
{
"id": "tbkH5dJGSU",
"initial_rating": 3,
"confidence": 4,
"soundness": 2,
"contribution": 2,
"presentation": 3,
"summary": "This paper proposes Hard View Pretraining (HVP), an approach for improving self-supervised learning (SSL) by selecting challenging, high-loss views during pretrai... | {
"rating": "3;5;5;6",
"rating_avg": 4.75,
"confidence": "4;4;3;5",
"confidence_avg": 4,
"soundness": "2;3;3;3",
"soundness_avg": 2.75,
"contribution": "2;2;3;2",
"contribution_avg": 2.25,
"presentation": "3;3;3;3",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.114082"
} | {
"id": "N3SgMdBy5F",
"metareview": "The rebuttal provided clarifications about the proposed method and its analysis that were useful for assessing the paper's contribution and responded adequately to most reviewer concerns. After discussion, reviewer AeuA recommended acceptance, k3VL recommended marginal acceptanc... | {
"decision": "Accept (Poster)"
} |
AKAz88zYLB | 2409.20412v1 | Conformal Prediction for Dose-Response Models with Continuous Treatments | {
"content": "## Abstract\n\nAbstract Understanding the dose-response relation between a continuous treatment and the outcome for an individual can greatly drive decision-making, particularly in areas like personalized drug dosing and personalized healthcare interventions. Point estimates are often insufficient in th... | [
{
"id": "xrRCPHKCb1",
"initial_rating": 8,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 4,
"summary": "The authors propose a conformal prediction-based method for estimating uncertainty in the dose-response function, which defines the effect of continuous treatment... | {
"rating": "3;5;5;6;8",
"rating_avg": 5.4,
"confidence": "4;3;2;3;4",
"confidence_avg": 3.2,
"soundness": "1;3;2;3;3",
"soundness_avg": 2.4,
"contribution": "1;1;2;2;3",
"contribution_avg": 1.8,
"presentation": "2;2;3;3;4",
"presentation_avg": 2.8
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:01.114715"
} | {
"id": "OB7W7DexzO",
"metareview": "This paper proposed a conformal prediction-based method for uncertainty quantification in dose-response models with continuous treatments. The approach leverages weighted conformal prediction, incorporating propensity estimation and kernel functions to address covariate shifts. ... | {
"decision": "Reject"
} |
ALzTQUgW8a | 2410.16179v2 | MagicPIG: LSH Sampling for Efficient LLM Generation | {
"content": "## Abstract\n\nAbstract Large language models (LLMs) with long context windows have gained significant attention. However, the KV cache, stored to avoid re-computation, becomes a bottleneck. Various dynamic sparse or TopK-based attention approximation methods have been proposed to leverage the common in... | [
{
"id": "OyHSYBG7sP",
"initial_rating": 6,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 4,
"summary": "This paper studies the optimization of long-context LLM inference. Unlike most existing approaches that mainly adopt TopK selection for attention calculation, thi... | {
"rating": "6;6;6;8;8",
"rating_avg": 6.8,
"confidence": "4;4;4;3;4",
"confidence_avg": 3.8,
"soundness": "3;3;3;3;3",
"soundness_avg": 3,
"contribution": "3;3;3;4;3",
"contribution_avg": 3.2,
"presentation": "3;2;4;4;3",
"presentation_avg": 3.2
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Spotlight",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.115409"
} | {
"id": "PoehvVuP7c",
"metareview": "This paper studies the optimization of long-context LLM inference and instead of using TopK selection for attention calculation, presents a method based on LSH and importance sampling. Experiments are encouraging.",
"additional_comments": "Rebuttal was satisfactory and helped ... | {
"decision": "Accept (Spotlight)"
} |
AMegoEnlpS | 2410.11551v1 | LoKO: Low-Rank Kalman Optimizer for Online Fine-Tuning of Large Models | {
"content": "## Abstract\n\nAbstract Training large models with millions or even billions of parameters from scratch incurs substantial computational costs. Parameter Efficient Fine-Tuning (PEFT) methods, particularly Low-Rank Adaptation (LoRA), address this challenge by adapting only a reduced number of parameters ... | [
{
"id": "4tsbLyNwrU",
"initial_rating": 3,
"confidence": 3,
"soundness": 3,
"contribution": 2,
"presentation": 2,
"summary": "Low-Rank Kalman Optimizer (LoKO) is a novel optimizer for Parameter-Efficient Fine-Tuning (PEFT) aimed specifically at LoRA. LoKO frames fine-tuning as an optimal... | {
"rating": "3;5;6;6",
"rating_avg": 5,
"confidence": "3;3;4;4",
"confidence_avg": 3.5,
"soundness": "3;3;4;3",
"soundness_avg": 3.25,
"contribution": "2;2;3;3",
"contribution_avg": 2.5,
"presentation": "2;3;3;3",
"presentation_avg": 2.75
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:01.116159"
} | {
"id": "S0bMJnsPi1",
"metareview": "## Summary\n\nLoKO is a novel optimization algorithm that integrates the Extended Kalman Filter (EKF) with Low-Rank Adaptation (LoRA) for online fine-tuning of large pre-trained models. It uses LoRA's low-rank decomposition to reduce the number of trainable parameters and a diag... | {
"decision": "Reject"
} |
AMkf7h7HER | 2405.15757v2 | Looking Backward: Streaming Video-to-Video Translation with Feature Banks | {
"content": "## Abstract\n\nAbstract This paper introduces StreamV2V, a diffusion model that achieves real-time streaming video-to-video (V2V) translation with user prompts.\nUnlike prior V2V methods using batches to process limited frames, we opt to process frames in a streaming fashion, to support unlimited frames... | [
{
"id": "woL1cJRxpF",
"initial_rating": 6,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "This paper presents StreamV2V, a real-time streaming video-to-video (V2V) translation method leveraging a backward-looking feature bank to maintain temporal consi... | {
"rating": "5;6;6;6;6;6;8",
"rating_avg": 6.142857142857143,
"confidence": "4;3;3;4;3;3;3",
"confidence_avg": 3.2857142857142856,
"soundness": "3;3;2;3;4;3;3",
"soundness_avg": 3,
"contribution": "2;4;3;3;3;3;3",
"contribution_avg": 3,
"presentation": "3;3;3;3;4;3;3",
"presentation_avg": 3.14285714... | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.116921"
} | {
"id": "oczIHXUZUy",
"metareview": "This paper proposed SteamV2V, a diffusion model that could do real-time streaming video-to-video translation with user prompts. Authors achieved this by maintaining a feature bank. The proposed solution is faster than previous work while maintaining temporal consistency.\n\n7 re... | {
"decision": "Accept (Poster)"
} |
AN6PIiObp0 | 2411.01798v1 | SALSA: Soup-based Alignment Learning for Stronger Adaptation in RLHF | {
"content": "## Abstract\n\nAbstract In Large Language Model (LLM) development, Reinforcement Learning from Human Feedback (RLHF) is crucial for aligning models with human values and preferences. RLHF traditionally relies on the Kullback-Leibler (KL) divergence between the current policy and a frozen initial policy ... | [
{
"id": "8tp4mShrlu",
"initial_rating": 5,
"confidence": 2,
"soundness": 2,
"contribution": 2,
"presentation": 3,
"summary": "The paper studies RLHF and proposes SALSA, which leverages a \"model soup\" approach, averaging weights from two supervised fine-tuned (SFT) models to create a mo... | {
"rating": "3;5;5;5;5;5",
"rating_avg": 4.666666666666667,
"confidence": "4;3;4;4;4;2",
"confidence_avg": 3.5,
"soundness": "3;3;3;2;3;2",
"soundness_avg": 2.6666666666666665,
"contribution": "2;2;3;2;2;2",
"contribution_avg": 2.1666666666666665,
"presentation": "3;3;2;3;3;3",
"presentation_avg": 2... | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:01.117620"
} | {
"id": "MVB1aDUohD",
"metareview": "This paper proposes a model-soup technique which uses the average weight to construct the KL center in RLHF. The majority of the reviewers vote for rejection. The main complains are missing sufficient, rigorous explanation of the phenomena and the current empirical finding has b... | {
"decision": "Reject"
} |
AOlm45AUVS | 2410.01101v1 | Exploiting Structure in Offline Multi-Agent RL: The Benefits of Low Interaction Rank | {
"content": "## Abstract\n\nAbstract We study the problem of learning an approximate equilibrium in the offline multi-agent reinforcement learning (MARL) setting. We introduce a structural assumption—the interaction rank —and establish that functions with low interaction rank are significantly more robust to distrib... | [
{
"id": "L4DgSDvkM6",
"initial_rating": 5,
"confidence": 4,
"soundness": 2,
"contribution": 2,
"presentation": 3,
"summary": "This paper proposes a new notion of interaction rank to characterize the structure of a Markov game. If the true reward function and the learned reward function h... | {
"rating": "5;6;6;8",
"rating_avg": 6.25,
"confidence": "4;3;2;3",
"confidence_avg": 3,
"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": "2026-01-14T22:16:01.118463"
} | {
"id": "lwpU7UHm3r",
"metareview": "All reviewers appreciate the novelty of interaction rank, and its tight statistical characterization in multi-agent RL. Accept.",
"additional_comments": "NA"
} | {
"decision": "Accept (Poster)"
} |
APCjgjFy5M | 2312.12339v2 | Value Explicit Pretraining for Learning Transferable Representations | {
"content": "## Abstract\n\nAbstract We propose Value Explicit Pretraining (VEP), a method that learns generalizable representations for transfer reinforcement learning. VEP enables learning of new tasks that share similar objectives as previously learned tasks, by learning an encoder for objective-conditioned repre... | [
{
"id": "Xc2thAM1JE",
"initial_rating": 5,
"confidence": 5,
"soundness": 3,
"contribution": 3,
"presentation": 2,
"summary": "This paper presents a novel transfer learning method. The aim is to learn generalizable representations from offline expert data that can be effectively transferr... | {
"rating": "3;3;3;5",
"rating_avg": 3.5,
"confidence": "3;5;3;5",
"confidence_avg": 4,
"soundness": "2;3;2;3",
"soundness_avg": 2.5,
"contribution": "2;2;2;3",
"contribution_avg": 2.25,
"presentation": "3;3;1;2",
"presentation_avg": 2.25
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:01.119425"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
APDnmucgID | 2402.10958v2 | Relative Preference Optimization: Enhancing LLM Alignment through Contrasting Responses across Identical and Diverse Prompts | {
"content": "## Abstract\n\nAbstract In the field of large language models (LLMs), aligning models with the diverse preferences of users is a critical challenge. Direct Preference Optimization (DPO) has played a key role in this area. It works by using pairs of preferences derived from the same prompts, and it funct... | [
{
"id": "HGskuAedej",
"initial_rating": 5,
"confidence": 5,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "The authors proposed new preference optimization methods to improve the performance of preference learning on various tasks. They explained that the current state... | {
"rating": "3;5;5;6",
"rating_avg": 4.75,
"confidence": "4;4;5;4",
"confidence_avg": 4.25,
"soundness": "3;2;3;3",
"soundness_avg": 2.75,
"contribution": "2;2;3;3",
"contribution_avg": 2.5,
"presentation": "3;3;3;3",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:01.120113"
} | {
"id": "xa9CRsrXuJ",
"metareview": "This paper proposes Relative Preference Optimization (RPO) which is designed to discern between responses derived from identical and related prompts, which enables utilizing both paired and unpaired data. While the reviewers find the method to be interesting, they also raised si... | {
"decision": "Reject"
} |
APojAzJQiq | 2408.11104v1 | ConFIG: Towards Conflict-free Training of Physics Informed Neural Networks | {
"content": "## Abstract\n\nAbstract The loss functions of many learning problems contain multiple additive terms that can disagree and yield conflicting update directions. For Physics-Informed Neural Networks (PINNs), loss terms on initial/boundary conditions and physics equations are particularly interesting as th... | [
{
"id": "3Yxlxu8Cd1",
"initial_rating": 8,
"confidence": 4,
"soundness": 4,
"contribution": 3,
"presentation": 4,
"summary": "Motivated by the discrepancy between the residual and data loss terms in Physics-Informed Neural Networks (PINNs), the authors propose a \"conflict-free\" gradien... | {
"rating": "5;6;8;8",
"rating_avg": 6.75,
"confidence": "4;3;4;3",
"confidence_avg": 3.5,
"soundness": "3;3;4;4",
"soundness_avg": 3.5,
"contribution": "3;3;4;3",
"contribution_avg": 3.25,
"presentation": "3;4;3;4",
"presentation_avg": 3.5
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Spotlight",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.120896"
} | {
"id": "Id9hjLVUqU",
"metareview": "This paper proposes **ConFIG** for multi-objective optimization, ensuring positive cosine similarity between update directions and consistent optimization rates for all loss terms. The authors also introduce a momentum-based variant to accelerate convergence. The paper presents ... | {
"decision": "Accept (Spotlight)"
} |
ARQIJXFcTH | 2309.16519v3 | AtomSurf: Surface Representation for Learning on Protein Structures | {
"content": "## Abstract\n\nAbstract While there has been significant progress in evaluating and comparing different representations for learning on protein data, the role of surface-based learning approaches remains not well-understood. In particular, there is a lack of direct and fair benchmark comparison between ... | [
{
"id": "4m7Y2dmP2A",
"initial_rating": 6,
"confidence": 5,
"soundness": 3,
"contribution": 3,
"presentation": 2,
"summary": "The paper aims to improve our understanding of surface-based learning for protein representations. This is achieved by first adapting the surface encoder from the... | {
"rating": "5;6;6;6",
"rating_avg": 5.75,
"confidence": "3;4;4;5",
"confidence_avg": 4,
"soundness": "3;4;3;3",
"soundness_avg": 3.25,
"contribution": "2;2;3;3",
"contribution_avg": 2.5,
"presentation": "3;4;4;2",
"presentation_avg": 3.25
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.121803"
} | {
"id": "5Ar0EymDsN",
"metareview": "The paper received positive support from most of the reviewers, and the overall recommendation was positive. Thus an accept is recommended.",
"additional_comments": "Most of the issues have been resolved, and the reviewer giving a score of 5 did not respond to author rebuttals... | {
"decision": "Accept (Poster)"
} |
AS8SPTyBgw | 2411.01992v1 | Turing completeness of prompting | {
"content": "## Abstract\n\nAbstract Since the success of GPT, large language models (LLMs) have been revolutionizing machine learning and have initiated the so-called LLM prompting paradigm. In the era of LLMs, people train a single general-purpose LLM and provide the LLM with different prompts to perform different... | [
{
"id": "qNOG5UxZRC",
"initial_rating": 6,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "This is a theoretical work whose main contribution is the proof that there exists a finite-sized transformer such that for any computable function $\\varphi$ ther... | {
"rating": "5;5;6;6",
"rating_avg": 5.5,
"confidence": "3;2;2;4",
"confidence_avg": 2.75,
"soundness": "4;3;3;3",
"soundness_avg": 3.25,
"contribution": "2;2;3;3",
"contribution_avg": 2.5,
"presentation": "3;3;3;3",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.122620"
} | {
"id": "pbQA2gfpaf",
"metareview": "This paper proves that LLM prompting is Turing complete, extending theoretical understanding to a common one-model-many-task setting of LLMs. I am not an expert in this area and lean on the reviewers for decision making. The reviewers appreciated the importance, novelty and tech... | {
"decision": "Accept (Poster)"
} |
AUBvo4sxVL | 2410.21317v1 | MatExpert: Decomposing Materials Discovery By Mimicking Human Experts | {
"content": "## Abstract\n\nAbstract Material discovery is a critical research area with profound implications for various industries. In this work, we introduce MatExpert , a novel framework that leverages Large Language Models (LLMs) and contrastive learning to accelerate the discovery and design of new solid-stat... | [
{
"id": "qMBv14xICE",
"initial_rating": 6,
"confidence": 3,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "MatExpert is designed to streamline the discovery of new materials using LLMs and contrastive learning. Inspired by the traditional workflow of human experts, Mat... | {
"rating": "3;5;5;6",
"rating_avg": 4.75,
"confidence": "3;4;3;3",
"confidence_avg": 3.25,
"soundness": "2;2;2;3",
"soundness_avg": 2.25,
"contribution": "2;2;3;2",
"contribution_avg": 2.25,
"presentation": "2;3;3;3",
"presentation_avg": 2.75
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.123522"
} | {
"id": "eRxuzEpSaI",
"metareview": "The paper introduces MatExpert, a framework that leverages Large Language Models (LLMs) and contrastive learning to streamline the discovery of new solid-state materials. By emulating expert workflows through retrieval, transition, and generation stages, MatExpert demonstrates s... | {
"decision": "Accept (Poster)"
} |
AUCYptvAf3 | 2410.03072v1 | Multi-Robot Motion Planning with Diffusion Models | {
"content": "## Abstract\n\nAbstract Diffusion models have recently been successfully applied to a wide range of robotics applications for learning complex multi-modal behaviors from data. However, prior works have mostly been confined to single-robot and small-scale environments due to the high sample complexity of... | [
{
"id": "LS6sJHkRSN",
"initial_rating": 6,
"confidence": 4,
"soundness": 4,
"contribution": 3,
"presentation": 4,
"summary": "This paper presents a multi-robot motion planning algorithm leveraging diffusion models and MAPF algorithms. In the proposed MMD algorithm, a conditional diffusio... | {
"rating": "6;6;6;8",
"rating_avg": 6.5,
"confidence": "4;4;4;3",
"confidence_avg": 3.75,
"soundness": "3;3;3;3",
"soundness_avg": 3,
"contribution": "3;3;3;4",
"contribution_avg": 3.25,
"presentation": "4;2;4;4",
"presentation_avg": 3.5
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Spotlight",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.124192"
} | {
"id": "5iYgfXIqPp",
"metareview": "This paper introduces the Multi-robot Multi-model planning Diffusion (MMD) method, which addresses a key challenge in multi-robot motion planning (MRMP) by leveraging single-robot diffusion models combined with Multi-Agent Path Finding (MAPF) constraints. The proposed approach s... | {
"decision": "Accept (Spotlight)"
} |
AV7OXVlAyi | 2410.04780v1 | Mitigating Modality Prior-Induced Hallucinations in Multimodal Large Language Models via Deciphering Attention Causality | {
"content": "## Abstract\n\nAbstract Multimodal Large Language Models (MLLMs) have emerged as a central focus in both industry and academia, but often suffer from biases introduced by visual and language priors, which can lead to multimodal hallucination. These biases arise from the visual encoder and the Large Lang... | [
{
"id": "CzaNkgbrRV",
"initial_rating": 3,
"confidence": 2,
"soundness": 2,
"contribution": 2,
"presentation": 2,
"summary": "The paper points out problems arising from biases induced by visual and language priors in the visual encoder and the LMM backbone, and it mentions the oversight ... | {
"rating": "3;3;5;5",
"rating_avg": 4,
"confidence": "3;2;4;5",
"confidence_avg": 3.5,
"soundness": "2;2;2;2",
"soundness_avg": 2,
"contribution": "2;2;2;3",
"contribution_avg": 2.25,
"presentation": "3;2;2;3",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.124809"
} | {
"id": "CmMMtXgOwQ",
"metareview": "### Summary:\nThis paper introduces CausalMM, a framework combining backdoor adjustment and counterfactual reasoning to mitigate hallucinations in multimodal large language models (MLLMs). The method treats modality priors as confounders between attention mechanisms and model ou... | {
"decision": "Accept (Poster)"
} |
AY89HCxunl | 2406.06449v2 | Cometh: A continuous-time discrete-state graph diffusion model | {
"content": "## Abstract\n\nAbstract The abstract paragraph should be indented 1/2 inch (3 picas) on both left and\nright-hand margins. Use 10 point type, with a vertical spacing of 11 points.\nThe word Abstract must be centered, in small caps, and in point size 12. Two\nline spaces precede the abstract. The abstrac... | [
{
"id": "IFtAAdT0ro",
"initial_rating": 5,
"confidence": 4,
"soundness": 3,
"contribution": 2,
"presentation": 2,
"summary": "This paper presents Cometh, a continuous-time discrete-state diffusion model for graph generation that combines the flexibility of continuous time with the struct... | {
"rating": "3;3;5;5",
"rating_avg": 4,
"confidence": "4;5;4;4",
"confidence_avg": 4.25,
"soundness": "3;3;3;3",
"soundness_avg": 3,
"contribution": "1;2;2;2",
"contribution_avg": 1.75,
"presentation": "3;3;3;2",
"presentation_avg": 2.75
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:01.125281"
} | {
"id": "BcTzOSHFiO",
"metareview": "**Summary**: This paper proposes a continuous-time discrete-state diffusion model for graphs, dubbed Cometh. Cometh combines the flexibility of continuous time with the structure-aware benefits of discrete-state modeling, with replacing structural encodings to a simple random-wa... | {
"decision": "Reject"
} |
AZTdO6JJKt | 2404.01965v2 | Leveraging AutoML for Sustainable Deep Learning: A Multi-Objective HPO Approach on Deep Shift Neural Networks | {
"content": "## Abstract\n\nAbstract Deep Learning (DL) has advanced various fields by extracting complex patterns from large datasets.\nHowever, the computational demands of DL models pose environmental and resource challenges. Deep shift neural networks (DSNNs) offer a solution by leveraging shift operations to re... | [
{
"id": "E9wHJmMJGM",
"initial_rating": 3,
"confidence": 5,
"soundness": 1,
"contribution": 1,
"presentation": 2,
"summary": "This paper investigates the potential of DSNNs to reduce computational complexity and environmental impact in deep learning (DL) models. Through AutoML techniques... | {
"rating": "3;5;5;5;6",
"rating_avg": 4.8,
"confidence": "5;3;3;3;3",
"confidence_avg": 3.4,
"soundness": "1;3;3;3;3",
"soundness_avg": 2.6,
"contribution": "1;2;2;2;2",
"contribution_avg": 1.8,
"presentation": "2;2;2;3;2",
"presentation_avg": 2.2
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:01.125996"
} | {
"id": "yw6cbiZiIL",
"metareview": "This paper presents an exploration into optimizing Deep Shift Neural Networks (DSNNs) for balancing performance and energy efficiency using Automated Machine Learning (AutoML) techniques. The authors propose a multi-fidelity, multi-objective hyperparameter optimization approach ... | {
"decision": "Reject"
} |
AbJWZp4THG | 2410.18117v1 | Efficient Adaptive Federated Optimization | {
"content": "## Abstract\n\nAbstract Adaptive optimization plays a pivotal role in federated learning, where simultaneous server and client-side adaptivity have been shown to be essential for maximizing its performance. However, the scalability of jointly adaptive systems is often constrained by limited resources in... | [
{
"id": "SaRA1cKNEv",
"initial_rating": 8,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "The paper proposes the algorithm FedAda^2 as a strategy for jointly training adaptive optimizers for both the client and server, but also saving on memory and com... | {
"rating": "3;3;5;6",
"rating_avg": 4.25,
"confidence": "4;3;3;3",
"confidence_avg": 3.25,
"soundness": "1;2;3;3",
"soundness_avg": 2.25,
"contribution": "1;2;3;2",
"contribution_avg": 2,
"presentation": "3;2;2;3",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:01.127102"
} | {
"id": "bPPKXa1pDb",
"metareview": "The paper presents results on adaptive optimization in the FL context, including the value of adaptive methods at the clients and efficient communication, say by not communicating the pre-conditioner, and also ideas on reducing on device memory. There is general agreement among ... | {
"decision": "Reject"
} |
AcAD4VEgCX | 2411.06525v1 | I2VControl-Camera: Precise Video Camera Control with Adjustable Motion Strength | {
"content": "## Abstract\n\nAbstract Video generation technologies are developing rapidly and have broad potential applications.\nAmong these technologies, camera control is crucial for generating professional-quality videos that accurately meet user expectations.\nHowever, existing camera control methods still suff... | [
{
"id": "JaJTUZONfA",
"initial_rating": 6,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 2,
"summary": "The submission pintroduces a new framework for training camer-controllable text-to-video model. To achieve the goal, the authors propose to disentangle the motion... | {
"rating": "5;5;6;8",
"rating_avg": 6,
"confidence": "3;2;4;3",
"confidence_avg": 3,
"soundness": "3;3;3;4",
"soundness_avg": 3.25,
"contribution": "2;2;3;4",
"contribution_avg": 2.75,
"presentation": "3;3;2;4",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.128573"
} | {
"id": "A0FX5PiVj6",
"metareview": "The paper presents a method for allowing more precise camera control in video generation. The reviewers appreciated the approach, which decomposes scene motion into global motion, attributed to camera movement and local motion induced by objects. This decomposition enables more ... | {
"decision": "Accept (Poster)"
} |
AcVpLS86RT | 2409.20558v1 | Uni$^2$Det: Unified and Universal Framework for Prompt-Guided Multi-dataset 3D Detection | {
"content": "## Abstract\n\nAbstract We present Uni 2 Det, a brand new framework for unified and universal multi-dataset training on 3D detection, enabling robust performance across diverse domains and generalization to unseen domains.\nDue to substantial disparities in data distribution and variations in taxonomy a... | [
{
"id": "QF3vYYLLn8",
"initial_rating": 8,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 4,
"summary": "This paper addresses a critical challenge in autonomous driving research: integrating multiple datasets collected across different locations and using varied sens... | {
"rating": "5;6;6;8",
"rating_avg": 6.25,
"confidence": "4;4;2;4",
"confidence_avg": 3.5,
"soundness": "2;3;3;3",
"soundness_avg": 2.75,
"contribution": "2;3;2;3",
"contribution_avg": 2.5,
"presentation": "2;2;3;4",
"presentation_avg": 2.75
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.129151"
} | {
"id": "LwAjNuSrLw",
"metareview": "This paper presents a novel unified framework for prompt-guided multi-dataset 3D detection. The proposed method shows very significant performance on several challenging benchmarks including KITTI, Waymo, and nuScenes. All the reviewers are positive about the technical contribut... | {
"decision": "Accept (Poster)"
} |
Acvo2RGSCy | 2402.02392v3 | DeLLMa: Decision Making Under Uncertainty with Large Language Models | {
"content": "## Abstract\n\nAbstract The potential of large language models (LLMs) as decision support tools is increasingly being explored in fields such as business, engineering, and medicine, which often face challenging tasks of decision-making under uncertainty .\nIn this paper, we show that directly prompting ... | [
{
"id": "4QpD4RwmHh",
"initial_rating": 6,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "The authors propose framework to work with LLMs for decision making under uncertainty; in doing so, they take classical utility/decision theoretic framework as ... | {
"rating": "5;6;8",
"rating_avg": 6.333333333333333,
"confidence": "4;4;4",
"confidence_avg": 4,
"soundness": "2;3;4",
"soundness_avg": 3,
"contribution": "2;3;3",
"contribution_avg": 2.6666666666666665,
"presentation": "2;3;3",
"presentation_avg": 2.6666666666666665
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Spotlight",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.129839"
} | {
"id": "UQ38Yf3cSV",
"metareview": "The reviewers were generally positive about the paper, specifically with the framework and the rigor in the writing. There are suggestions toc make the paper self-contained and not refer so much to the appendix, increase the details on the experiments and justification of the in... | {
"decision": "Accept (Spotlight)"
} |
AdiNf568ne | 2410.02760v1 | Erasing Conceptual Knowledge from Language Models | {
"content": "## Abstract\n\nAbstract Concept erasure in language models has traditionally lacked a comprehensive evaluation framework, leading to incomplete assessments of effectiveness of erasure methods. We propose an evaluation paradigm centered on three critical criteria: innocence (complete knowledge removal), ... | [
{
"id": "SbTW4ttgIA",
"initial_rating": 5,
"confidence": 4,
"soundness": 2,
"contribution": 2,
"presentation": 2,
"summary": "In this paper, the authors proposed to evaluate LLM concept erasure with innocence, seamlessness and specificity, and put forward Erasure of Language Memory (ELM)... | {
"rating": "3;5;5",
"rating_avg": 4.333333333333333,
"confidence": "5;3;4",
"confidence_avg": 4,
"soundness": "2;3;2",
"soundness_avg": 2.3333333333333335,
"contribution": "3;3;2",
"contribution_avg": 2.6666666666666665,
"presentation": "3;3;2",
"presentation_avg": 2.6666666666666665
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:01.130499"
} | {
"id": "Q7n5H1a7ri",
"metareview": "This submission obtains three negative rating. Several limitations of this version are still existed according to the results like limited experiments and non-significant improvements, undesirable effects of full finetuning, and more comparison with baselines. After rebuttal, tw... | {
"decision": "Reject"
} |
AepP8ddd3L | 2402.07812v2 | Retrieval Augmented Thought Process for Private Data Handling in Healthcare | {
"content": "## Abstract\n\nAbstract Large Language Models (LLMs) have demonstrated the strong potential to assist both clinicians and the general public with their extensive medical knowledge. However, their application in healthcare is constrained due to concerns about the privacy of data used in training, which p... | [
{
"id": "aZQUnOTlPi",
"initial_rating": 5,
"confidence": 3,
"soundness": 2,
"contribution": 2,
"presentation": 3,
"summary": "Large Language Models (LLMs) two major challenges in healthcare such as:\n1. Privacy concerns with sensitive medical data\n2. Difficulty in reliably accessing and... | {
"rating": "3;3;3;5;6",
"rating_avg": 4,
"confidence": "4;4;4;3;3",
"confidence_avg": 3.6,
"soundness": "2;1;2;2;3",
"soundness_avg": 2,
"contribution": "2;2;2;2;2",
"contribution_avg": 2,
"presentation": "2;3;1;3;2",
"presentation_avg": 2.2
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:01.131164"
} | {
"id": "FmkYJNOGn1",
"metareview": "The paper introduces Retrieval-Augmented Thought Process (RATP), leveraging Monte Carlo Tree Search to enhance privacy-preserving reasoning in healthcare, achieving superior accuracy and robustness against imperfect retrieval in question-answering tasks.\n\nStrengths\n- RATP com... | {
"decision": "Reject"
} |
AfSNOjtWyt | 2407.03310v1 | Universal length generalization with Turing Programs | {
"content": "## Abstract\n\nAbstract Length generalization refers to the ability to extrapolate from short training sequences to long test sequences and is a challenge for current large language models. While prior work has proposed some architecture or data format changes to achieve length generalization, these pro... | [
{
"id": "CzuJga5Wcf",
"initial_rating": 5,
"confidence": 4,
"soundness": 3,
"contribution": 2,
"presentation": 2,
"summary": "Authors showed that a (very long) CoT which slightly change an input $x_i$ into $x_{i+1}$ up to a point where it solves a problem completely can achieve state-of-... | {
"rating": "3;5;6",
"rating_avg": 4.666666666666667,
"confidence": "3;4;2",
"confidence_avg": 3,
"soundness": "2;3;4",
"soundness_avg": 3,
"contribution": "2;2;2",
"contribution_avg": 2,
"presentation": "2;2;4",
"presentation_avg": 2.6666666666666665
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:01.131842"
} | {
"id": "mC4lKLRZCO",
"metareview": "This paper proposes \"Turing Programs\", a Chain-of-Thought strategy mimicking Turing Machine computations to improve length generalization in transformers for algorithmic tasks. The authors claim universality, demonstrating generalization on addition, multiplication, and in-co... | {
"decision": "Reject"
} |
AfZH9EEuRR | 2410.05497v1 | EgoQR: Efficient QR Code Reading in Egocentric Settings | {
"content": "## Abstract\n\nAbstract QR codes have become ubiquitous in daily life, enabling rapid information exchange. With the increasing adoption of smart wearable devices, there is a need for efficient, and friction-less QR code reading capabilities from Egocentric point-of-views. However, adapting existing pho... | [
{
"id": "skSmtUF47v",
"initial_rating": 1,
"confidence": 5,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "This paper proposes a QR code reader for wearable devices. The Implementation of this system incorporates existing modules such as Faster-RCNN and ZXing. In addit... | {
"rating": "1;1;3;3;3",
"rating_avg": 2.2,
"confidence": "4;5;5;3;3",
"confidence_avg": 4,
"soundness": "2;3;1;2;3",
"soundness_avg": 2.2,
"contribution": "1;2;1;2;2",
"contribution_avg": 1.6,
"presentation": "1;3;1;2;3",
"presentation_avg": 2
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:01.132433"
} | {
"id": "ZrETbnyqI3",
"metareview": "The paper introduces EgoQR, a system for efficient QR code reading on smart wearable devices in egocentric settings. It addresses challenges like code distortion, wide field-of-view, and resource constraints with a detection component using Faster R-CNN and decoding enhanced by ... | {
"decision": "Reject"
} |
Ah3n8U3kRT | 2402.02461v4 | Median Clipping for Zeroth-order Non-Smooth Convex Optimization and Multi Arm Bandit Problem with Heavy-tailed Symmetric Noise | {
"content": "## Abstract\n\nAbstract In this paper, we consider non-smooth convex optimization with a zeroth-order oracle corrupted by symmetric stochastic noise. Unlike the existing high-probability results requiring the noise to have bounded κ 𝜅 \\kappa italic_κ -th moment with κ ∈ ( 1 , 2 ] 𝜅 1 2 \\kappa\\in(1,... | [
{
"id": "8ytCNMJE5Q",
"initial_rating": 6,
"confidence": 3,
"soundness": 4,
"contribution": 3,
"presentation": 3,
"summary": "This paper consider non-smooth convex optimization with a zeroth-order oracle with heavy-tail symmetric stochastic noise. Algorithms are proposed for settings wit... | {
"rating": "5;5;6;6",
"rating_avg": 5.5,
"confidence": "3;4;3;3",
"confidence_avg": 3.25,
"soundness": "3;3;3;4",
"soundness_avg": 3.25,
"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": "2026-01-14T22:16:01.133252"
} | {
"id": "3qIojf8Jlf",
"metareview": "This paper studies zeroth-order non-smooth convex optimization with heavy-tailed symmetric noise. Compared to previous algorithms, the algorithm proposed in this paper could handle heavier noise, i.e., noise with bounded $\\kappa$-th moment for any $\\kappa>0$. The authors furth... | {
"decision": "Reject"
} |
AhcYq4CnfF | 2410.06549v1 | DiffGAD: A Diffusion-based Unsupervised Graph Anomaly Detector | {
"content": "## Abstract\n\nAbstract Graph Anomaly Detection (GAD) is crucial for identifying abnormal entities within networks, garnering significant attention across various fields. Traditional unsupervised methods, which decode encoded latent representations of unlabeled data with a reconstruction focus, often fa... | [
{
"id": "l9qlypBU2P",
"initial_rating": 5,
"confidence": 3,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "The paper proposes a Diffusion-based Graph Anomaly Detector for unsupervised identifying abnormal entities in networks. The contributions can be summarized as fol... | {
"rating": "3;5;6;6;8",
"rating_avg": 5.6,
"confidence": "5;3;3;3;3",
"confidence_avg": 3.4,
"soundness": "2;3;3;3;3",
"soundness_avg": 2.8,
"contribution": "2;2;2;3;3",
"contribution_avg": 2.4,
"presentation": "2;3;2;3;3",
"presentation_avg": 2.6
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.134259"
} | {
"id": "ENNNP9OLT7",
"metareview": "This paper proposes a diffusion-based graph anomaly detector (DiffGAD) to identify abnormal entities in networks. DiffGAD encodes graph data into a latent space via an encoder, adds noise and samples using unconditional and conditional diffusion models, and transformers embeddi... | {
"decision": "Accept (Poster)"
} |
Ahlrf2HGJR | 2402.15449v1 | Repetition Improves Language Model Embeddings | {
"content": "## Abstract\n\nAbstract Recent approaches to improving the extraction of text embeddings from autoregressive large language models (LLMs) have largely focused on improvements to data, backbone pretrained language models, or improving task-differentiation via instructions. In this work, we address an arc... | [
{
"id": "iRtmnjcAEh",
"initial_rating": 6,
"confidence": 3,
"soundness": 4,
"contribution": 3,
"presentation": 4,
"summary": "The paper proposes a very easy method to improve the language model embedding. They introduce echo embedding by repeating the input and extracting embedding from ... | {
"rating": "6;6;6;6",
"rating_avg": 6,
"confidence": "4;4;4;3",
"confidence_avg": 3.75,
"soundness": "3;3;2;4",
"soundness_avg": 3,
"contribution": "3;3;3;3",
"contribution_avg": 3,
"presentation": "3;3;3;4",
"presentation_avg": 3.25
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.135049"
} | {
"id": "GTQ2iFDGnL",
"metareview": "The paper proposes “echo embedding,” a method to improve language model embeddings by repeating the input and extracting embeddings from the repeated tokens. It aims to address the limitations of causal attention in autoregressive models for text embedding tasks without architec... | {
"decision": "Accept (Poster)"
} |
AjXkRZIvjB | 2410.05229v1 | GSM-Symbolic: Understanding the Limitations of Mathematical Reasoning in Large Language Models | {
"content": "## Abstract\n\nAbstract Recent advancements in Large Language Models (LLMs) have sparked interest in their formal reasoning capabilities, particularly in mathematics. The GSM8K benchmark is widely used to assess the mathematical reasoning of models on grade-school-level questions. While the performance ... | [
{
"id": "4wTQfanAL0",
"initial_rating": 8,
"confidence": 4,
"soundness": 3,
"contribution": 4,
"presentation": 2,
"summary": "This paper creates a novel dataset (GSM-Symbolic) with several different variants to test whether LLMs are capable of generalized reasoning. GSM-Symbolic has seve... | {
"rating": "3;5;6;8",
"rating_avg": 5.5,
"confidence": "4;4;4;4",
"confidence_avg": 4,
"soundness": "2;3;3;3",
"soundness_avg": 2.75,
"contribution": "2;2;3;4",
"contribution_avg": 2.75,
"presentation": "2;2;4;2",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.135744"
} | {
"id": "JKK28GQRjJ",
"metareview": "The paper introduces GSM-Symbolic, a novel benchmark designed to evaluate the mathematical reasoning capabilities of large language models (LLMs) with a focus on controlled and nuanced assessments. Building on the widely used GSM8K benchmark, GSM-Symbolic employs symbolic templa... | {
"decision": "Accept (Poster)"
} |
AjunxrcKa2 | 2408.01415v1 | Conditional LoRA Parameter Generation | {
"content": "## Abstract\n\nAbstract Generative models have achieved remarkable success in image, video, and text domains.\nInspired by this, researchers have explored utilizing generative models to generate neural network parameters.\nHowever, these efforts have been limited by the parameter size and the practicali... | [
{
"id": "4epzg9hUlX",
"initial_rating": 1,
"confidence": 5,
"soundness": 2,
"contribution": 1,
"presentation": 1,
"summary": "The authors propose a LoRA parameter diffusion models conditioned on specific image and NLP tasks.",
"strengths": "1. The presented qualitative results are on... | {
"rating": "1;3;3;5;5",
"rating_avg": 3.4,
"confidence": "5;4;4;4;3",
"confidence_avg": 4,
"soundness": "2;2;2;3;2",
"soundness_avg": 2.2,
"contribution": "1;2;2;3;3",
"contribution_avg": 2.2,
"presentation": "1;2;1;3;2",
"presentation_avg": 1.8
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:01.136460"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
AkUer8ooMi | 2410.14138v1 | ProReason: Multi-Modal Proactive Reasoning with Decoupled Eyesight and Wisdom | {
"content": "## Abstract\n\nAbstract Large vision-language models (LVLMs) have witnessed significant progress on visual understanding tasks.\nHowever, they often prioritize language knowledge over image information on visual reasoning tasks, incurring performance degradation.\nTo tackle this issue, we first identify... | [
{
"id": "ZKRAjSlMT3",
"initial_rating": 6,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "This paper discusses the performance degradation problem of Large Vision-language models (LVLMs), which tend to rely more on language knowledge than image informa... | {
"rating": "5;6;6",
"rating_avg": 5.666666666666667,
"confidence": "5;5;4",
"confidence_avg": 4.666666666666667,
"soundness": "2;3;3",
"soundness_avg": 2.6666666666666665,
"contribution": "2;2;3",
"contribution_avg": 2.3333333333333335,
"presentation": "2;3;3",
"presentation_avg": 2.666666666666666... | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:01.137089"
} | {
"id": "yeufIfMOI4",
"metareview": "Paper addresses degradation of the performance of VLMs that stem from them relying less on visual information and more on language priors. The proposed approach amounts to Chain-of-Reasoning with multiple agents, which, in-part decouple visual information gathering and reasoning... | {
"decision": "Reject"
} |
Alba3Y7hcs | 2410.10998v1 | WILT: A Multi-Turn, Memorization-Robust Inductive Logic Benchmark for LLMs | {
"content": "## Abstract\n\nAbstract While large language models (LLMs) have shown impressive capabilities across a wide range of domains, they still encounter significant challenges in reasoning tasks that require gathering evidence over multiple turns and drawing logical conclusions from this evidence. These chall... | [
{
"id": "4V8lVwyT3a",
"initial_rating": 6,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "The paper introduces WILT (Wason Inductive Logic Test), a new reasoning benchmark for LLMs to evaluate reasoning over multiple turns.\n\nInspired by the Wason 2-4... | {
"rating": "3;3;5;6",
"rating_avg": 4.25,
"confidence": "4;3;4;4",
"confidence_avg": 3.75,
"soundness": "2;3;3;3",
"soundness_avg": 2.75,
"contribution": "2;2;2;3",
"contribution_avg": 2.25,
"presentation": "2;2;3;3",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:01.138181"
} | {
"id": "vq35AQzY59",
"metareview": "The paper introduces WILT (Wason Inductive Logic Test), a new benchmark designed to evaluate the multi-turn reasoning capabilities of Large Language Models (LLMs). Inspired by the classic Wason 2-4-6 task, WILT requires LLMs to infer a hidden logical or arithmetic function of th... | {
"decision": "Reject"
} |
AlsvUVZFE9 | 2410.08549v2 | SCORE NEURAL OPERATOR: A GENERATIVE MODEL FOR LEARNING AND GENERALIZING ACROSS MULTIPLE PROBABILITY DISTRIBUTIONS | {
"content": "## Abstract\n\nAbstract Most existing generative models are limited to learning a single probability distribution from the training data and cannot generalize to novel distributions for unseen data.\nAn architecture that can generate samples from both trained datasets and unseen probability distribution... | [
{
"id": "XKYp5c647X",
"initial_rating": 5,
"confidence": 3,
"soundness": 3,
"contribution": 2,
"presentation": 1,
"summary": "The authors propose an improved score based model, capable of generalizing to unseen continuous distributions that are given as an input at test time. The distrib... | {
"rating": "1;3;5;5;5",
"rating_avg": 3.8,
"confidence": "4;3;3;2;3",
"confidence_avg": 3,
"soundness": "2;2;3;2;3",
"soundness_avg": 2.4,
"contribution": "2;2;2;2;2",
"contribution_avg": 2,
"presentation": "1;1;3;1;1",
"presentation_avg": 1.4
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:01.139160"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
AmEgWDhmTr | 2410.05459v1 | From Sparse Dependence to Sparse Attention: Unveiling How Chain-of-Thought Enhances Transformer Sample Efficiency | {
"content": "## Abstract\n\nAbstract Chain-of-thought (CoT) significantly enhances the reasoning performance of large language models (LLM). While current theoretical studies often attribute this improvement to increased expressiveness and computational capacity, we argue that expressiveness is not the primary limit... | [
{
"id": "dFGbULhlqW",
"initial_rating": 8,
"confidence": 4,
"soundness": 4,
"contribution": 4,
"presentation": 4,
"summary": "The paper studies how Transformers may or may not solve k-sparse parity problems. Without chain of thought, the authors provide a lower bound for one-pass algorit... | {
"rating": "5;5;8;8",
"rating_avg": 6.5,
"confidence": "4;4;3;4",
"confidence_avg": 3.75,
"soundness": "2;2;3;4",
"soundness_avg": 2.75,
"contribution": "3;3;4;4",
"contribution_avg": 3.5,
"presentation": "2;4;4;4",
"presentation_avg": 3.5
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.140159"
} | {
"id": "NqfA1okUhk",
"metareview": "This paper investigates how Chain-of-Thought (CoT) enhances the sample efficiency of Transformer models in reasoning tasks, focusing on the parity learning problem. The authors theoretically demonstrate an exponential gap in sample complexity between learning with and without Co... | {
"decision": "Accept (Poster)"
} |
AnL6BuWzxa | 2410.03052v1 | Learning Structured Representations by Embedding Class Hierarchy with Fast Optimal Transport | {
"content": "## Abstract\n\nAbstract To embed structured knowledge within labels into feature representations, prior work (Zeng et al., 2022 ) proposed to use the C o p henetic C orrelation C oefficient (CPCC) as a regularizer during supervised learning. This regularizer calculates pairwise Euclidean distances of cl... | [
{
"id": "BRGaRlttHH",
"initial_rating": 5,
"confidence": 3,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "The paper aims to improve classification by incorporating similarities among classes. These semantic relationships between labels are introduced into the original... | {
"rating": "5;5;6;6;6",
"rating_avg": 5.6,
"confidence": "3;3;4;2;3",
"confidence_avg": 3,
"soundness": "3;3;3;3;3",
"soundness_avg": 3,
"contribution": "2;2;3;3;3",
"contribution_avg": 2.6,
"presentation": "3;3;3;3;3",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.141315"
} | {
"id": "FU6oV6lOrI",
"metareview": "$\\ell_2$-CPCC has been used to embed structured knowledge within labels into feature representations, but has the limitation that it may misrepresent class relationships, especially in multi-modal cases. To address this limitation, the paper propose a EMD-CPCC that use earth mo... | {
"decision": "Accept (Poster)"
} |
Antib6Uovh | 2403.02233v2 | A Theoretical Analysis of Self-Supervised Learning for Vision Transformers | {
"content": "## Abstract\n\nAbstract Masked reconstruction, which predicts randomly masked patches from unmasked ones, has emerged as an important approach in self-supervised pretraining. However, the theoretical understanding of masked pretraining is rather limited, especially for the foundational architecture of t... | [
{
"id": "JiHsPwUeet",
"initial_rating": 5,
"confidence": 3,
"soundness": 2,
"contribution": 3,
"presentation": 1,
"summary": "This paper is about understanding and contrasting the behavior of two pretraining techniques used for ViT architectures: Masked Auto-encoding and Contrastive Lear... | {
"rating": "5;5;6;8",
"rating_avg": 6,
"confidence": "3;3;2;4",
"confidence_avg": 3,
"soundness": "3;2;3;4",
"soundness_avg": 3,
"contribution": "3;3;3;4",
"contribution_avg": 3.25,
"presentation": "2;1;3;3",
"presentation_avg": 2.25
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.142663"
} | {
"id": "1IowAXpFNT",
"metareview": "Summary: \nThis paper presents a theoretical analysis of self-supervised learning for Vision Transformers (ViTs), focusing on the differences between contrastive learning (CL) and masked autoencoders (MAE). By analyzing the training dynamics of one-layer softmax-based self-atte... | {
"decision": "Accept (Poster)"
} |
AqHbMV28o7 | 2405.15941v1 | A Unified Theory of Stochastic Proximal Point Methods without Smoothness | {
"content": "## Abstract\n\nAbstract This paper presents a comprehensive analysis of a broad range of variations of the stochastic proximal point method ( SPPM ). Proximal point methods have attracted considerable interest owing to their numerical stability and robustness against imperfect tuning, a trait not shared... | [
{
"id": "lSgACDw4po",
"initial_rating": 3,
"confidence": 5,
"soundness": 3,
"contribution": 2,
"presentation": 4,
"summary": "The paper deals with variance reduction techniques for stochastic proximal point algorithms, which recently emerged as a robust alternative to stochastic gradient... | {
"rating": "5;5;5",
"rating_avg": 5,
"confidence": "4;3;4",
"confidence_avg": 3.6666666666666665,
"soundness": "2;2;3",
"soundness_avg": 2.3333333333333335,
"contribution": "2;2;2",
"contribution_avg": 2,
"presentation": "3;2;3",
"presentation_avg": 2.6666666666666665
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:01.144667"
} | {
"id": "EjD0d4Ds1Y",
"metareview": "A framework of convergence analysis is presented for a range of variations of SPPM, such as variance reduction and Point-SAGA. While the analysis is sound, reviewers assess that the novelty is limited and the contribution is a little overstated.\n\nExperiments are limited to the... | {
"decision": "Reject"
} |
AqfUa08PCH | 2410.02749v2 | Training Language Models on Synthetic Edit Sequences Improves Code Synthesis | {
"content": "## Abstract\n\nAbstract Software engineers mainly write code by editing existing programs. In contrast, large language models (LLMs) autoregressively synthesize programs in a single pass. One explanation for this is the scarcity of open-sourced edit data. While high-quality instruction data for code syn... | [
{
"id": "Mz97MoNyWG",
"initial_rating": 5,
"confidence": 3,
"soundness": 2,
"contribution": 3,
"presentation": 3,
"summary": "This paper proposes LintSeq, which is a synthetic data generation algorithm that generates edit sequences from raw code to construct training datasets for LLMs. S... | {
"rating": "5;5;5;5",
"rating_avg": 5,
"confidence": "5;5;4;3",
"confidence_avg": 4.25,
"soundness": "2;3;2;2",
"soundness_avg": 2.25,
"contribution": "2;2;3;3",
"contribution_avg": 2.5,
"presentation": "3;1;1;3",
"presentation_avg": 2
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.146658"
} | {
"id": "M6nB4MnP6n",
"metareview": "This paper presents LintSeq, an algorithm that refactors existing code into sequences of structured edits for training language models. The key strengths include: (1) Novel approach to generate synthetic edit sequences that preserve syntax validity using linters, (2) Comprehensi... | {
"decision": "Accept (Poster)"
} |
AqueuvXErD | 2403.06013v1 | Are Classification Robustness and Explanation Robustness Really Strongly Correlated? An Analysis Through Input Loss Landscape | {
"content": "## Abstract\n\nAbstract This paper delves into the critical area of deep learning robustness, challenging the conventional belief that classification robustness and explanation robustness in image classification systems are inherently correlated. Through a novel evaluation approach leveraging clustering... | [
{
"id": "r47Nm3Tmdt",
"initial_rating": 3,
"confidence": 4,
"soundness": 1,
"contribution": 2,
"presentation": 3,
"summary": "This paper examines the concept of explanation robustness through the lens of classification robustness in the context of targeted adversarial attacks. While clas... | {
"rating": "3;3;3;5",
"rating_avg": 3.5,
"confidence": "2;4;4;3",
"confidence_avg": 3.25,
"soundness": "2;2;1;3",
"soundness_avg": 2,
"contribution": "2;2;2;2",
"contribution_avg": 2,
"presentation": "2;2;3;1",
"presentation_avg": 2
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:01.147575"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
ArwsbHBoxA | 2404.10776v1 | Nearly Optimal Algorithms for Contextual Dueling Bandits from Adversarial Feedback | {
"content": "## Abstract\n\nAbstract Learning from human feedback plays an important role in aligning generative models, such as large language models (LLM). However, the effectiveness of this approach can be influenced by adversaries, who may intentionally provide misleading preferences to manipulate the output in ... | [
{
"id": "8sJXPqQUj3",
"initial_rating": 5,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 2,
"summary": "This paper studies the contextual dueling bandit problem with adversarial preference feedback in which adversaries may intentionally provide misleading preference... | {
"rating": "5;5;5;6",
"rating_avg": 5.25,
"confidence": "3;4;3;3",
"confidence_avg": 3.25,
"soundness": "3;3;3;3",
"soundness_avg": 3,
"contribution": "2;3;3;3",
"contribution_avg": 2.75,
"presentation": "3;3;2;3",
"presentation_avg": 2.75
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:01.148666"
} | {
"id": "kqlFrHzJ2O",
"metareview": "This paper examines the contextual dueling bandit problem with potentially adversarial feedback and proposes a robust algorithm. The algorithm is shown to achieve a regret upper bound that depends on the total amount of adversarial feedback, $C$. The main weaknesses include the ... | {
"decision": "Reject"
} |
AsAy7CROLs | 2305.12883v3 | Prediction Risk and Estimation Risk of the Ridgeless Least Squares Estimator under General Assumptions on Regression Errors | {
"content": "## Abstract\n\nAbstract. In recent years, there has been a significant growth in research focusing on minimum ℓ 2 subscript ℓ 2 \\ell_{2} roman_ℓ start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT norm (ridgeless) interpolation least squares estimators. However, the majority of these analyses have been limited to ... | [
{
"id": "982fWHlmHb",
"initial_rating": 6,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "In this paper, the authors investigate prediction risk and estimation risk under more general regression error assumptions beyond i.i.d. errors. In particlar, th... | {
"rating": "3;6;8",
"rating_avg": 5.666666666666667,
"confidence": "4;4;3",
"confidence_avg": 3.6666666666666665,
"soundness": "3;3;4",
"soundness_avg": 3.3333333333333335,
"contribution": "1;3;3",
"contribution_avg": 2.3333333333333335,
"presentation": "3;3;4",
"presentation_avg": 3.33333333333333... | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:01.149786"
} | {
"id": "ZwhowOJWrY",
"metareview": "This paper studies both the prediction risk as well as the estimation risk in non-trivial minimum l2 norm or 'ridgeless' least squares problems. In particular, they consider a general setting that leads to a better understanding of non-iid settings, including a relationship betw... | {
"decision": "Accept (Poster)"
} |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.