{"total":12,"items":[{"citing_arxiv_id":"2606.26396","ref_index":49,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"At the Edge of Understanding: Sparse Autoencoders Trace The Limits of Transformer Generalization","primary_cat":"cs.LG","submitted_at":"2026-06-24T21:26:43+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":4.0,"formal_verification":"none","one_line_summary":"Sparse autoencoders show OOD prompts increase fallacious concept activation in transformers, offering a mechanistic measure of shift and a path to robust fine-tuning.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2606.12138","ref_index":4,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Unstable Features, Reproducible Subspaces: Understanding Seed Dependence in Sparse Autoencoders","primary_cat":"cs.LG","submitted_at":"2026-06-10T14:32:57+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"Stable SAE features dominate functional signal while unstable features concentrate in reproducible subspaces, allowing more stable SAEs via cross-seed pooling.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2606.03002","ref_index":14,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Perplexity Can Miss SAE Feature Damage Under Quantization","primary_cat":"cs.LG","submitted_at":"2026-06-02T01:17:05+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"Quantization of LLMs can degrade many SAE features even when perplexity improves or stays similar, as shown by correlation measurements on frozen SAEs for Pythia-70M and Gemma-2-2B models across INT8 to INT4.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.28149","ref_index":23,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Sign-Aware Gated Sparse Autoencoders: Modeling Anticorrelated Features with Bi-Jump-ReLU Activations","primary_cat":"cs.LG","submitted_at":"2026-05-27T08:31:43+00:00","verdict":null,"verdict_confidence":null,"novelty_score":null,"formal_verification":null,"one_line_summary":null,"context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.18229","ref_index":26,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Are Sparse Autoencoder Benchmarks Reliable?","primary_cat":"cs.LG","submitted_at":"2026-05-18T11:20:57+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"An audit of SAEBench reveals that Targeted Probe Perturbation and Spurious Correlation Removal metrics fail reliability tests and should not be used to evaluate sparse autoencoders.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.14347","ref_index":39,"ref_count":2,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Exemplar Partitioning for Mechanistic Interpretability","primary_cat":"cs.LG","submitted_at":"2026-05-14T04:15:30+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":7.0,"formal_verification":"none","one_line_summary":"Exemplar Partitioning creates Voronoi partitions of LLM activation space via leader clustering on streamed activations, yielding comparable, interpretable dictionaries that support interventions and achieve competitive benchmark results with ~1000x less compute than SAEs.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.12991","ref_index":26,"ref_count":2,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Not Just RLHF: Why Alignment Alone Won't Fix Multi-Agent Sycophancy","primary_cat":"cs.LG","submitted_at":"2026-05-13T04:45:08+00:00","verdict":null,"verdict_confidence":null,"novelty_score":null,"formal_verification":null,"one_line_summary":null,"context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.12874","ref_index":23,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Descriptive Collision in Sparse Autoencoder Auto-Interpretability: When One Explanation Describes Many Features","primary_cat":"cs.LG","submitted_at":"2026-05-13T01:41:38+00:00","verdict":"ACCEPT","verdict_confidence":"MODERATE","novelty_score":8.0,"formal_verification":"none","one_line_summary":"Many distinct SAE features share identical explanations, with the average annotation resolving only 70% of feature identity in a large annotated dataset.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.12770","ref_index":33,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"WriteSAE: Sparse Autoencoders for Recurrent State","primary_cat":"cs.LG","submitted_at":"2026-05-12T21:32:45+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":8.0,"formal_verification":"none","one_line_summary":"WriteSAE introduces sparse autoencoders with rank-1 matrix atoms for recurrent state updates, allowing replacement tests that outperform deletion on 92.4% of positions and a formula predicting logit changes with R²=0.98.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.04072","ref_index":15,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Sparse Autoencoder Decomposition of Clinical Sequence Model Representations: Feature Complexity, Task Specialisation, and Mortality Prediction","primary_cat":"cs.LG","submitted_at":"2026-04-13T12:08:32+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":7.0,"formal_verification":"none","one_line_summary":"Sparse autoencoders applied to a 14.5M-parameter clinical EHR model reveal progressive abstraction across layers, with SAE features outperforming dense ones for mortality in full-sequence probes but not in leakage-safe windows where dense representations match or exceed them.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2512.06655","ref_index":13,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Graph-Regularized Sparse Autoencoders for LLM Safety Steering","primary_cat":"cs.LG","submitted_at":"2025-12-07T04:46:30+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"GSAE improves selective refusal on safety benchmarks by smoothing SAE directions over a co-activation graph and applying them via a two-gate controller, outperforming standard SAEs and baselines on Llama-3 and other models.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2509.09708","ref_index":24,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Beyond I'm Sorry, I Can't: Dissecting Large Language Model Refusal","primary_cat":"cs.CL","submitted_at":"2025-09-07T02:29:07+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"Sparse autoencoders plus greedy filtering and factorization-machine interaction modeling identify minimal sets of features in Gemma-2-2B-IT and LLaMA-3.1-8B-IT whose ablation produces jailbreaks by flipping refusal to compliance.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null}],"limit":50,"offset":0}