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
54 results for “keyword:"inference efficiency"” in ICLR 2027. 21 withdrawn or desk rejected listed apart. clearsort:relevanceratingnewestmost read
[PDF]HAIR: Fine-Strand Hidden-State Anchored Implicit Reasoning with Single-Prefill Inference▲3.0k2.40
[PDF]Many Caches, Same Reasoning: Rethinking Importance Selection for Reasoning-Time KV Compression▼1.2k1.67
[PDF]When to Wait Before Fan-Out: Warm-Up, Verification, and Residency in Cache-Aware LLM Evaluation▲9.2k2.67
[PDF]ARMAD: Adaptive Routing for Heterogeneous Multi-Agent Debate in Large Language Model Reasoning▼1.2k1.67
[PDF]PRISM: Progressive Resolution with Independent Section Mapping for Scalable User Profiling▲11.00
[PDF]SelKV: Selective KV Cache Merging with Per-Token Merge-or-Drop and Attention Compensation▲5.7k2.00
[PDF]Keys Decide Attention, Values Decide Contribution: Value-Gated KV Cache Eviction at 20× Compression▼1351.33
[PDF]AgSpec: Pushing the Limits of Retrieval-Based Speculative Decoding in Coding Agent Pipelines▼4.5k2.67
[PDF]Content Before Position: A Bayesian Decision Framework for Diffusion Language Model Decoding▼5.8k2.33
[PDF]Prune Less, Pay More: Imperceptible Budget Inflation Attacks on Token Pruning in Vision-Language Models▼1.2k1.67
[PDF]When Does Adaptive Test-Time Augmentation Pay Off? A Matched Execution Study in Dense Vision▼1.2k1.67
[PDF]Where Does Layer-Skip Fragility Live? A Matched-Harness Decomposition of Training-Free Adaptive Layer Skipping▲11.00
[PDF]Look Back to Think Less: Rethinking Visual Token Pruning Efficiency under Multimodal Reasoning Cost▼1.2k3.00
Withdrawn or desk rejected (21)
[PDF]Cross-Model KV Cache Construction in LLM Familiies: Residual-Stream Mapping Without Matched Head GeometryDesk rejected
[PDF]Commit Order, Not Commit Rate: Answer Availability and Prefix Preservation in Diffusion Language ModelsWithdrawn1.67
[PDF]SurroGate: Efficient Text-to-Image Routing via Continuous Generative Profiling in Sparse MoEsDesk rejected
[PDF]VeloCache: Frequency-Decoupled Velocity Caching for Training-Free Video Diffusion AccelerationWithdrawn2.00
[PDF]Grammar Reachability for Lossless LM-Head Compression and Deterministic Token Injection in Structured GenerationWithdrawn1.33
[PDF]Cross-dLLM: Cross-Block Diffusion Decoding with Autoregressive Drafting for Complex ReasoningDesk rejected
[PDF]-VLM: Detecting and Correcting Vision-Language Hallucination via Active Representational Perturbation and Residual Spectral GroundingWithdrawn1.33