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
52 results for “keyword:"ablation"” in ICLR 2027. 14 withdrawn or desk rejected listed apart. clearsort:relevanceratingnewestmost read
[PDF]Human Difficulty Is Not Task Difficulty: What Policy Representations Know Beyond an Optimal Solver▼1.2k1.67
[PDF]Four Cells Are Not the Same Dose: Size-Conditioned Spatial Lesions in Detection Transformers▲3.0k2.33
[PDF]From Behavior to Mechanism: Tracing Divergent Response Modes in Frontier Language Models▼1351.33
[PDF]Testing Attribution Claims in Diffusion-LM Remasking: A Localized Effect Without a Demonstrated Benefit▼5.3k1.33
[PDF]Anchored but Not Protected: When Refusal Erodes During Fine-Tuning, and Whether Holding One Direction Stops It▲13k2.33
[PDF]Path-Space: Estimating Transformer Component Effects and Their Interactions from Shared Backward Passes▼5.8k2.33
[PDF]The Saussurean Representation Hypothesis: A Causal Account of Representational Structures in LLMs▼4.3k2.00
[PDF]Finding Structure Without Fitting Noise: Taking Apart the Small-Table Advantage of Tabular Foundation Models▼4.3k2.00
[PDF]Does Register-Length Flexibility Help Peptide–MHC-II Classification? A Controlled Ablation on the Chicken Allele BLB2▼1351.33
[PDF]Spoken-Digit Recognition Without Training: Geometry, Coupling, and Drive Effects in Oscillator Networks▼1.2k1.67
[PDF]Interventional Commutativity: Effect Preservation and Reference Selection Under Weight-Only Quantisation▼1.2k1.67
[PDF]Hidden Equivalences in Sheaf Neural Networks: The Share Credited to Matrix-Valued Transport Is Not Intrinsic▲1.7k3.33
[PDF]Beyond Average Contribution: Class-Conditional Modality-Missingness Signatures Across Emotion Datasets, Models, and Humans▼5.8k2.33
[PDF]Saturated Screening, Concentrated Cost: What a Full-Stack Ablation Table Reads in a Redundant Refusal Stack▼1.2k1.67
[PDF]Attention Tracks Reaction Centers, But Doesn't Need Them: A Causal Study of a Retrosynthesis Transformer▼4.3k2.00
[PDF]CROSS-TASK GAINS WITHOUT AUXILIARY LABELS: ARCHITECTURE-PRESERVING ABLATIONS UNDER DISJOINT SUPERVISION▼1.2k1.67
[PDF]Most of the Ablation Gain Was the Baseline's Training Budget: A Placebo-Controlled Audit of Evidence-Guided Deepfake Detection▼5.8k2.33
[PDF]Only the Story Completes the Grid: Hand-Authored and Grammar-Generated Puzzles for Narrative Augmented Reasoning▲20k2.67
[PDF]LOCA-RCA: A Label-Free Counterfactual Benchmark for Dependence on Cited Evidence in RCA Agents▼4.3k2.00
[PDF]Sin-SMFL: An Ablation Study of Modulation, Class Balancing, and Gain Calibration in Detection Losses▼4.3k2.00
Withdrawn or desk rejected (14)
[PDF]Propagation Is Not Decoration: When Physical Diffraction Carries Generative ComputationWithdrawn1.00
[PDF]Token-Level Trust Regions under Reuse: Component Effects, Stability, and InterventionDesk rejected
[PDF]Structure Before Attention: A Zero-Initialized Convolution Recovers Windowed Language-Model QualityWithdrawn1.33
[PDF]Which Image Property Carries The Jailbreak? A Controlled Dissection Of Image-To-Text JailbreaksWithdrawn1.67
[PDF]Better Behavioral Prediction, More Faithful Model Ablations? Evidence from Sequential ChoiceWithdrawn1.33