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
31 results for “keyword:"length generalization"” in ICLR 2027. 5 withdrawn or desk rejected listed apart. clearsort:relevanceratingnewestmost read
[PDF]SUCCINCTNESS IS NOT IDENTIFIABILITY: FINITE-EVIDENCE LIMITS FOR TRANSFORMER LENGTH GEN-ERALIZATION▲11.00
[PDF]An Analysis of Position Encoding in Hybrid LLMs With Local Mixing and Global NoPE Attention▼4.5k2.67
[PDF]Not All Thinking is Created Equal: Latent Reasoning Discovers a Recurrent Search Algorithm for Depth Generalization▲6.1k2.00
[PDF]Train Short, Act Long: Multistability and the Horizon Independence of Recurrent Policies▼1.2k1.67
[PDF]Beyond the Training Horizon: Mechanisms and Limits of Length Generalization in Looped Transformers▲13k2.33
[PDF]Quantitative Generalization Bounds for Multi-Layer Transformers: A Few Long Strings can Beat Many Short Ones▲153.33
Withdrawn or desk rejected (5)
[PDF]A Unified Formal Framework To Predict Length Generalization in Linear RNNs, Transformers and Hybrid ModelsWithdrawn2.00
[PDF]What the Loss Never Asks For: Which Group a Recurrent Network Represents, and Which Words Break ItDesk rejected