{"slug":"interpretability","topic":"interpretability","title":"Mechanistic Interpretability: State of the Field (October 2026)","summary":"An expert briefing on mechanistic interpretability as of late October 2026: superposition and linear features, sparse autoencoders and transcoders, automated circuit discovery and attribution graphs, causal-abstraction theory, probing, activation steering and the use of these tools in model audits. The central takeaway is that the tools now trace parts of the computation of production-scale models and appear in pre-deployment audits, yet controlled benchmarks find sparse-dictionary features often no better than neurons, linear probes or prompting on practical tasks, and the faithfulness and uniqueness of explanations remain unresolved. Work in 2026 concentrates on automating attribution-graph analysis, ground-truth testbeds and formal guarantees.","version":1,"published_at":"2026-10-29","updated_at":"2026-10-30","stats":{"words":6205,"citations":108,"claims":29,"timeline":19,"benchmarks":15,"open_problems":7,"key_papers":15,"glossary":19},"verification":{"status":"verified","method":"automated figure and citation-relevance triage against cited abstracts, targeted AI review with web checks of flagged statements, and a close-out pass hedging or removing unconfirmed statements; automated citation resolution","claims_checked":29,"issues_fixed":14,"verified_at":"2026-10-30"},"ai_assisted":true,"price":{"usd":"0.20","protocol":"x402","network":"eip155:84532","asset":"USDC"},"url":"https://space-kitty.com/dossiers/interpretability","preview_url":"https://space-kitty.com/api/v1/dossiers/interpretability/preview","purchase_url":"https://space-kitty.com/api/v1/dossiers/interpretability","toc":[{"level":2,"text":"Executive summary","anchor":"executive-summary"},{"level":2,"text":"Background and core concepts","anchor":"background-and-core-concepts"},{"level":2,"text":"Timeline of key milestones","anchor":"timeline-of-key-milestones"},{"level":2,"text":"State of the art","anchor":"state-of-the-art"},{"level":3,"text":"Superposition and feature geometry","anchor":"superposition-and-feature-geometry"},{"level":3,"text":"Sparse autoencoders: scale, variants and pathologies","anchor":"sparse-autoencoders-scale-variants-and-pathologies"},{"level":3,"text":"Transcoders, replacement models and attribution graphs","anchor":"transcoders-replacement-models-and-attribution-graphs"},{"level":3,"text":"Circuit discovery and causal localization","anchor":"circuit-discovery-and-causal-localization"},{"level":3,"text":"Probing, latent knowledge and lenses","anchor":"probing-latent-knowledge-and-lenses"},{"level":3,"text":"Steering and representation engineering","anchor":"steering-and-representation-engineering"},{"level":3,"text":"Auditing, model diffing and monitoring","anchor":"auditing-model-diffing-and-monitoring"},{"level":3,"text":"Beyond language models","anchor":"beyond-language-models"},{"level":3,"text":"Interpretable-by-design models and parameter-space methods","anchor":"interpretable-by-design-models-and-parameter-space-methods"},{"level":3,"text":"Quantitative benchmarks","anchor":"quantitative-benchmarks"},{"level":2,"text":"Open problems and active debates","anchor":"open-problems-and-active-debates"},{"level":2,"text":"Major programs and players","anchor":"major-programs-and-players"},{"level":2,"text":"Outlook: what to watch (November 2026 to April 2028)","anchor":"outlook-what-to-watch-november-2026-to-april-2028"},{"level":2,"text":"Key papers","anchor":"key-papers"},{"level":2,"text":"Glossary","anchor":"glossary"}],"sample_claims":["Superposition remains the organizing hypothesis for polysemantic neurons: networks store more sparse features than they have dimensions and accept interference between them.","Sparse autoencoders trained on language-model activations yield feature directions that automated methods judge more interpretable and monosemantic than directions from alternative decompositions."],"includes":["markdown","claims","timeline","benchmarks","open_problems","key_papers","glossary","bibliography"],"citation_count":108,"how_to_buy":"GET https://space-kitty.com/api/v1/dossiers/interpretability with an x402 client ($0.20, USDC on Base Sepolia (testnet))."}