ml/nlp phding @ usc, currently visiting harvard; training & interpretability & reasoning iglee.mehttps://bsky.app/profile/wordscompute.bsky.socialhttps://bsky.app/profile/wordscompute.bsky.social/post/3mobrjt6oak2zreally excited to head home for icml:) and attending the co-located FAR.ai alignment workshop (for the first time)! would love to meet others interested in training & interpretability http://FAR.ai14 Jun 2026 22:03 +0000at://did:plc:o2bxof3c7t77ltc7eahqrnt3/app.bsky.feed.post/3mobrjt6oak2zhttps://bsky.app/profile/wordscompute.bsky.social/post/3mobrejwb2j2gBenchmarks can be superficial, but model explanations and evaluations are fundamentally intertwined. What if we used interpretability as principled, scientific evaluation? If it met scientific standards? arxiv.org/abs/2605.05508 coming to EvalEval at ACL as oral 🧵 1/614 Jun 2026 22:00 +0000at://did:plc:o2bxof3c7t77ltc7eahqrnt3/app.bsky.feed.post/3mobrejwb2j2ghttps://bsky.app/profile/wordscompute.bsky.social/post/3mely7aw3tl2vNew dataset 🗂️ coming to #eacl What is (correct) reasoning in LLMs? How do you rigorously define/measure process fidelity? How might we study its acquisition in large scale training? We made a gigantic, verifiably correct reasoning traces of first order logic expressions! 1/911 Feb 2026 17:17 +0000at://did:plc:o2bxof3c7t77ltc7eahqrnt3/app.bsky.feed.post/3mely7aw3tl2vhttps://bsky.app/profile/wordscompute.bsky.social/post/3melxwcv2cz2tone of my new years "considerations" is to be less silent #onhere. so i guess i'll be #here and maybe also #there til february 15th11 Feb 2026 17:12 +0000at://did:plc:o2bxof3c7t77ltc7eahqrnt3/app.bsky.feed.post/3melxwcv2cz2thttps://bsky.app/profile/wordscompute.bsky.social/post/3melxnhdliz2tReally excited to receive Coefficient Giving's Technical AI Safety Research Grant via Berkeley Existential Risk Initiative w/ @nsaphra.bsky.social! We aim to predict potential AI model failures before impact--before deployment, using interpretability.11 Feb 2026 17:07 +0000at://did:plc:o2bxof3c7t77ltc7eahqrnt3/app.bsky.feed.post/3melxnhdliz2thttps://bsky.app/profile/wordscompute.bsky.social/post/3lormn65n6k24weve reached that point in this submission cycle, no amount of coffee will do 😞🙂‍↔️😞09 May 2025 23:51 +0000at://did:plc:o2bxof3c7t77ltc7eahqrnt3/app.bsky.feed.post/3lormn65n6k24https://bsky.app/profile/wordscompute.bsky.social/post/3llik77gmzs23INCOMING [contains quote post or other embedded content]29 Mar 2025 04:58 +0000at://did:plc:o2bxof3c7t77ltc7eahqrnt3/app.bsky.feed.post/3llik77gmzs23https://bsky.app/profile/wordscompute.bsky.social/post/3llik6mtmyk23titled: peer review29 Mar 2025 04:58 +0000at://did:plc:o2bxof3c7t77ltc7eahqrnt3/app.bsky.feed.post/3llik6mtmyk23https://bsky.app/profile/wordscompute.bsky.social/post/3lkeyopiask27really excited to be headed to OFC in SF! so excited to revisit optical physics 😀15 Mar 2025 01:42 +0000at://did:plc:o2bxof3c7t77ltc7eahqrnt3/app.bsky.feed.post/3lkeyopiask27https://bsky.app/profile/wordscompute.bsky.social/post/3lgo5hpcfyc22it’s finally raining in la:)26 Jan 2025 19:20 +0000at://did:plc:o2bxof3c7t77ltc7eahqrnt3/app.bsky.feed.post/3lgo5hpcfyc22https://bsky.app/profile/wordscompute.bsky.social/post/3lfcug2wffs2ii go on a really long walk almost every day, and at a high point in silverlake, i saw fire from all sides. and it's harder to breathe. and everything is orange.09 Jan 2025 14:14 +0000at://did:plc:o2bxof3c7t77ltc7eahqrnt3/app.bsky.feed.post/3lfcug2wffs2i