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
11,753 results for “keyword:"models"” in ICLR 2027. 2,918 withdrawn or desk rejected listed apart. clearsort:relevanceratingnewestmost read
[PDF]From Sparse to Dense: Diagnosing Task-Equivalent Language Robustness in Vision-Language-Action Models▲3.0k2.33
[PDF]HyperQuant: A Near-Entropy-Optimal Lattice Quantization Pipeline for Large Language and Diffusion Models▼5.9k2.33
[PDF]AliasGuard: Surrogate-Based Defense Against Passive Inference in Split Language Model Fine-Tuning▼4.3k2.00
[PDF]Memory-Gated Distillation: Mitigating Self-Generated Memory Drift in Autoregressive Diffusion World Model Distillation▼1.3k1.75
[PDF]Frames Don't Add Up: A Factorial Causal Audit of Contextual Evidence Utility in Video-Language Models▼1361.33
[PDF]Blame Errant Words, Not the Canvas: Defect-Aware Spatial Credit Assignment for Visual Text Generation▲5093.00
[PDF]Reference-Image-Guided Relation Editing via Token Bottlenecking and Semantic Shift Alignment▲3862.67
[PDF]Exploring the Boundary of Sparse Attention: Retrievability, Head Stability, and ANN Approximation▲9.2k2.67
Withdrawn or desk rejected (2,918)
[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