{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/transcoders-find-interpretable-llm-feature","title":"Transcoders Find Interpretable LLM Feature Circuits","arxiv_id":"2406.11944","date":"2024-06-17","proceeding":null,"authors":["Jacob Dunefsky","Philippe Chlenski","Neel Nanda"],"abstract":"A key goal in mechanistic interpretability is circuit analysis: finding sparse subgraphs of models corresponding to specific behaviors or capabilities. However, MLP sublayers make fine-grained circuit analysis on transformer-based language models difficult. In particular, interpretable features -- such as those found by sparse autoencoders (SAEs) -- are typically linear combinations of extremely many neurons, each with its own nonlinearity to account for. Circuit analysis in this setting thus either yields intractably large circuits or fails to disentangle local and global behavior. To address this we explore transcoders, which seek to faithfully approximate a densely activating MLP layer with a wider, sparsely-activating MLP layer. We introduce a novel method for using transcoders to perform weights-based circuit analysis through MLP sublayers. The resulting circuits neatly factorize into input-dependent and input-invariant terms. We then successfully train transcoders on language models with 120M, 410M, and 1.4B parameters, and find them to perform at least on par with SAEs in terms of sparsity, faithfulness, and human-interpretability. Finally, we apply transcoders to reverse-engineer unknown circuits in the model, and we obtain novel insights regarding the \"greater-than circuit\" in GPT2-small. Our results suggest that transcoders can prove effective in decomposing model computations involving MLPs into interpretable circuits. Code is available at https://github.com/jacobdunefsky/transcoder_circuits/.","url_abs":"https://arxiv.org/abs/2406.11944v2","url_pdf":"https://arxiv.org/pdf/2406.11944v2.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"transcoders-find-interpretable-llm-feature","repo_url":"https://github.com/jacobdunefsky/transcoder_circuits","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":0,"framework":"jax","reach":{"status":"ok"}}],"tasks":[],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":null,"atlas_url":"https://app.syntology.ai/?focus=2406.11944","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2406.11944"}},"developers":"https://syntology.ai/developers","read_at":"2026-09-24T18:15:14+00:00","read_at_is":"when the build read Syntology's graph, not when any sample ran","claim":"Per-sample execution status on synthesized fixtures; not a correctness claim about the paper. Samples come from repositories linked to the paper, official or community; repo_kind says which.","repos":[{"provenance":"deterministic:regex_extraction","url":"https://github.com/jacobdunefsky/transcoder_circuits","reach":{"status":"ok"}}],"summary":{"ran":6,"unverified":1},"by_repo_kind":{"official":{"samples":7,"ran":6,"repositories":1}},"repo_kind_vocabulary":{"official":"The archive marks this repository official for the paper","named_in_paper":"The archive records that the paper mentions this repository; it is not marked official","listed":"In the archive's code links for this paper, not marked official and not recorded as mentioned in the paper","found_in_text":"Syntology found this repository in the paper's own text; whether it is the authors' implementation is not asserted","community":"Not in the archive's code links for this paper; a community repository Syntology harvested"},"n_pointer_only_for_licence":7,"samples":[{"code_sha256_prefix":"7743fdc98d0dc90e","entry":"geometric_median_array","repo":"jacobdunefsky/transcoder_circuits","repo_kind":"official","path":"sae_training/geom_median/src/geom_median/numpy/weiszfeld_array.py","file_url":"https://github.com/jacobdunefsky/transcoder_circuits/blob/HEAD/sae_training/geom_median/src/geom_median/numpy/weiszfeld_array.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"7743fdc98d0dc90e"}},{"code_sha256_prefix":"736db4bd4f1d805e","entry":"geometric_median_per_component","repo":"jacobdunefsky/transcoder_circuits","repo_kind":"official","path":"sae_training/geom_median/src/geom_median/numpy/weiszfeld_array.py","file_url":"https://github.com/jacobdunefsky/transcoder_circuits/blob/HEAD/sae_training/geom_median/src/geom_median/numpy/weiszfeld_array.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"736db4bd4f1d805e"}},{"code_sha256_prefix":"109b275bea3f94e9","entry":"get_recons_loss","repo":"jacobdunefsky/transcoder_circuits","repo_kind":"official","path":"sae_training/train_sae_on_language_model.py","file_url":"https://github.com/jacobdunefsky/transcoder_circuits/blob/HEAD/sae_training/train_sae_on_language_model.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"109b275bea3f94e9"}},{"code_sha256_prefix":"72e1b397fe5aa9b2","entry":"get_scheduler","repo":"jacobdunefsky/transcoder_circuits","repo_kind":"official","path":"sae_training/optim.py","file_url":"https://github.com/jacobdunefsky/transcoder_circuits/blob/HEAD/sae_training/optim.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"72e1b397fe5aa9b2"}},{"code_sha256_prefix":"4daba25365b65503","entry":"sample_uniform","repo":"jacobdunefsky/transcoder_circuits","repo_kind":"official","path":"transcoder_circuits/feature_dashboards.py","file_url":"https://github.com/jacobdunefsky/transcoder_circuits/blob/HEAD/transcoder_circuits/feature_dashboards.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"4daba25365b65503"}},{"code_sha256_prefix":"0c1b8e7c5079b759","entry":"weighted_average","repo":"jacobdunefsky/transcoder_circuits","repo_kind":"official","path":"sae_training/geom_median/src/geom_median/numpy/weiszfeld_array.py","file_url":"https://github.com/jacobdunefsky/transcoder_circuits/blob/HEAD/sae_training/geom_median/src/geom_median/numpy/weiszfeld_array.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":true,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"0c1b8e7c5079b759"}},{"code_sha256_prefix":"0e24bc34a9a2f735","entry":"sample_percentiles","repo":"jacobdunefsky/transcoder_circuits","repo_kind":"official","path":"transcoder_circuits/feature_dashboards.py","file_url":"https://github.com/jacobdunefsky/transcoder_circuits/blob/HEAD/transcoder_circuits/feature_dashboards.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"0e24bc34a9a2f735"}}]},"arxiv_metadata":null,"syntology_extracted_results":null}