{"podcast":{"title":"LessWrong (Curated & Popular)","slug":"lesswrong-curated-popular-5643401","podcast_index_feed_id":5643401,"rss_url":"https://rss.buzzsprout.com/2037297.rss","website_url":"https://sites.libsyn.com/421877","image_url":"https://storage.buzzsprout.com/xq8g0aka74ttwoa9xkxlmy5fn9oa?.jpg","author":"LessWrong","episode_count":859,"summary":"Audio narrations of LessWrong posts. Includes all curated posts and all posts with 125+ karma. If you'd like more, subscribe to the “Lesswrong (30+ karma)” feed.","last_synced_at":null,"page_url":"https://stenobird.com/podcast/lesswrong-curated-popular-5643401"},"episode":{"title":"\"Natural Language Autoencoders Produce Unsupervised Explanations of LLM Activations\" by Subhash Kantamneni, kitft, Euan Ong, Sam Marks","slug":"natural-language-autoencoders-produce-unsupervised-explanations-of-llm-activations-by-subhash-kantamneni-kitft-euan-ong-sam-marks","published_at":"2026-05-08T05:45:49+00:00","page_url":"https://stenobird.com/podcast/lesswrong-curated-popular-5643401/natural-language-autoencoders-produce-unsupervised-explanations-of-llm-activations-by-subhash-kantamneni-kitft-euan-ong-sam-marks","show_page_url":"https://stenobird.com/podcast/lesswrong-curated-popular-5643401","url":"https://www.buzzsprout.com/2037297/episodes/19144572-natural-language-autoencoders-produce-unsupervised-explanations-of-llm-activations-by-subhash-kantamneni-kitft-euan-ong-sam-marks.mp3","audio_url":"https://www.buzzsprout.com/2037297/episodes/19144572-natural-language-autoencoders-produce-unsupervised-explanations-of-llm-activations-by-subhash-kantamneni-kitft-euan-ong-sam-marks.mp3","summary":"Abstract We introduce Natural Language Autoencoders (NLAs), an unsupervised method for generating natural language explanations of LLM activations. An NLA consists of two LLM modules: an activation verbalizer (AV) that maps an activation to a text description and an activation reconstructor (AR) that maps the description back to an activation. We jointly train the AV and AR with reinforcement learning to reconstruct residual stream activations. Although we optimize for activation reconstru...","meta_description":"Abstract We introduce Natural Language Autoencoders (NLAs), an unsupervised method for generating natural language explanations of LLM activations. An NLA…","key_points":[],"chapters":[],"topics":[],"duration_seconds":1092,"processing_state":"not_requested","actions":[{"name":"request_transcript","method":"POST","url":"https://stenobird.com/v1/public/podcasts/lesswrong-curated-popular-5643401/episodes/natural-language-autoencoders-produce-unsupervised-explanations-of-llm-activations-by-subhash-kantamneni-kitft-euan-ong-sam-marks/transcription-requests","description":"Idempotently request low-priority transcript generation for this episode."},{"name":"read_markdown","method":"GET","url":"https://stenobird.com/podcast/lesswrong-curated-popular-5643401/natural-language-autoencoders-produce-unsupervised-explanations-of-llm-activations-by-subhash-kantamneni-kitft-euan-ong-sam-marks.md","description":"Read the agent-friendly Markdown representation of this episode resource."}]}}