NeurIPS 2026 reviews are up: reviews, scores, rebuttals and meta-reviews for 7,877 papers. See the highest rated
acceptodds: the papers of ICLR 2027
11,743 results for “keyword:"models"” in ICLR 2027. 2,928 withdrawn or desk rejected listed apart. clearsort:relevanceratingnewestmost read
[PDF]Is your uncertainty map wrong, or is its target? Exact diagnostics for the Tweedie diagonal, and a gradient-free alternative▼1.2k1.67
[PDF]Timestep Weighting in Frozen Diffusion Classifiers: Global Partitions and Pointwise Reachability▼1.2k1.67
[PDF]HoltSeer: Stable Feature Forecasting under Sparse Anchors for Accelerated Diffusion Sampling▲3352.67
[PDF]Frozen Models, Evolving Expertise: Model-Agnostic Learning from Deployment Experience for Multimodal Medical AI▼4.3k2.00
[PDF]Fork-NFT: Forked-Credit Negative-Aware Fine-Tuning for Autoregressive Joint Audio–Video Generation▲3352.67
[PDF]ControlAnything: Unified word-level, Intensity-Aware, and Composable Control for Expressive Text-to-Speech▲3352.67
[PDF]Mitigating Cross-Modal Hallucinations by Correcting Modality Misrouting in Omnimodal LLMs▲3.1k2.33
[PDF]Faultless Misreading: Diagnosing Professional Agents Under Incomplete Spreadsheet Requests▲4.7k3.00
[PDF]Evidence-Aligned Multimodal On-Policy Self-Distillation for Fine-Grained Visual Understanding▼4.3k2.00
[PDF]Actions Speak Louder than Thoughts: State-Grounded On-Policy Distillation for Multi-Turn Agents▲9.3k2.67
[PDF]Detecting Time Series Anomalies Like an Expert: A Multi-Agent LLM Framework with Specialized Analyzers▼4.3k2.00
[PDF]Can we improve the performance of sparse attention in Grouped Query Experts without increasing experts?▲31.00
[PDF]Beyond Correctness: Benchmarking Response Behaviors and Improving Non-Thinking Outputs in Hybrid-Thinking MLLMs▲1.7k3.33
[PDF]Hallucinate, Filter, and Refine: Occluded Surface Reconstruction with Generative Priors▼4.3k2.00
Withdrawn or desk rejected (2,928)
[PDF]Mind Your Marginals: What Governs Fidelity in Linearized Hierarchical Optimal Transport?Desk rejected1.00
[PDF]A Window Is Not a Door: Mapping the Limits and Conditions of Knowledge Editing in Language ModelsDesk rejected2.33
[PDF]Propagation Is Not Decoration: When Physical Diffraction Carries Generative ComputationWithdrawn1.00
[PDF]When Is a Policy Violation a Model Failure? Policy Observability and Failure Attribution in Agent-Safety EvaluationDesk rejected