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
416 results for “keyword:"efficient inference"” in ICLR 2027. 79 withdrawn or desk rejected listed apart. clearsort:relevanceratingnewestmost read
[PDF]BridgeComp: Prompt Compression via Reasoning-Conditioned Attention and Anchor-Conditioned Policy Optimization▼1.2k3.00
[PDF]What Rank Calibration Actually Measures: The Sparse, Repeatable, Query-Specific Response of Video LLMs to KV-Cache Compression▼4.5k2.67
[PDF]LLMForge: Multi-Backend Hardware-Aware Neural Architecture Search with Infinite-Head Attention for Edge Language Models▼4.3k2.00
[PDF]Compile2Decide: Distilling Frontier Reasoning Models into Fast Calibrated Decision Models▼4.3k2.00
[PDF]SCOPE-Q: ONE-CHECKPOINT QUANTIZATION-AWARE TRAINING FOR CROSS-FRAMEWORK LLM DEPLOYMENT▼4.3k2.00
[PDF]Ask the Masked Block: Predicting Denoised-State Relevance for KV Cache Eviction in Diffusion Language Models▼5.8k2.33
[PDF]Seeing Action-Critical Evidence: Token Pruning with Internal Action Priors for Vision-Language-Action Models▲153.33
[PDF]Deeper Is Not Better for Quantized CLIP: Early Readout Recovers Accuracy While Cutting Compute▼5.8k2.33
[PDF]SCOPE: Self-Calibrated Ordered Principal Encoding for Tabular Foundation Models in High-Dimensional Data▼5.8k2.33
[PDF]DART: Decoupled Axis-Adaptive Reconstruction Trees for Budgeted Video Token Compression▼5.8k2.33
[PDF]TMOT: Geometric Optimal Transport over Target Representations for Speculative Verification▼1361.33
[PDF]XiYOLO: Energy-Aware Architecture Search for Object Detection from Sparse Hardware Measurements▼4.3k2.00
[PDF]ESPTQ: Post-Training Ternarization of Weights and KV Cache for Eliminating Per-Element Multiplications▼4.3k2.00
[PDF]PRISM: Principal-Residual Activation Sparsity via Structure-Aware Budget Allocation for Efficient LLM Inference▼5.8k2.33
[PDF]ReState: Enabling Cross-Query KV Reuse for Training-Free Acceleration of BERT Rerankers▲5103.00
[PDF]Where to Look, What to Keep: Two-Phase Learning for Grounding-Aware Visual Token Selection▲3972.67
Withdrawn or desk rejected (79)
[PDF]CCQ: Compensation-Conditioned Quadratic Selection for Absorbing Skipped Experts in MoE LLMsWithdrawn2.00
[PDF]VOLT: Joint Token Selection and Frame Budget Allocation via Subspace Volume MaximizationDesk rejected2.00
[PDF]ReCo:Training-Free Visual Token Pruning via Relational Transition Coreset in Large Vision-Language ModelsWithdrawn2.00
[PDF]DepthPruner: Visual Token Pruning with Pluggable Depth Guidance for Efficient Vision-Language-Action InferenceWithdrawn2.00