PhD candidate at LMU Munich. Representations, model and data attribution, training dynamics. Strong opinions on coffee and tea ☕ https://florian-eichin.comhttps://bsky.app/profile/florian-eichin.comhttps://bsky.app/profile/florian-eichin.com/post/3lumzyt6xrs2ySome recommendations for #ACL2025 👇 (join me and @janetlauyeung.bsky.social to talk about discourse generalization and probing!) [contains quote post or other embedded content]23 Jul 2025 12:38 +0000at://did:plc:molt37ru3i4ryruglxltcv3b/app.bsky.feed.post/3lumzyt6xrs2yhttps://bsky.app/profile/florian-eichin.com/post/3ltmdzr5ub22mPaper alert 🛎️ [contains quote post or other embedded content]10 Jul 2025 12:40 +0000at://did:plc:molt37ru3i4ryruglxltcv3b/app.bsky.feed.post/3ltmdzr5ub22mhttps://bsky.app/profile/florian-eichin.com/post/3lruz3etbqc2qCan you point to where "LLM knowledge" can be found in this graph? Source: en.wikipedia.org/wiki/Knowledge18 Jun 2025 12:28 +0000at://did:plc:molt37ru3i4ryruglxltcv3b/app.bsky.feed.post/3lruz3etbqc2qhttps://bsky.app/profile/florian-eichin.com/post/3lqfil75etc2yWant to know if your prompting is also affected by this? Addressing this and other issues systematically, we proposed Spotlight, which utilizes data mining to uncover the effects of prompt- and model-changes (meet us at ACL to discuss) arxiv.org/abs/2504.15815 [contains quote post or other embedded content]30 May 2025 14:57 +0000at://did:plc:molt37ru3i4ryruglxltcv3b/app.bsky.feed.post/3lqfil75etc2yhttps://bsky.app/profile/florian-eichin.com/post/3lqfgtevgo22yHappy not Friday everyone [contains quote post or other embedded content]30 May 2025 14:26 +0000at://did:plc:molt37ru3i4ryruglxltcv3b/app.bsky.feed.post/3lqfgtevgo22yhttps://bsky.app/profile/florian-eichin.com/post/3lqafnmuc6s2tProposing "kilometers cycled in 2 weeks" as a new success metric for research groups 🚴‍♀️📊🥇 @mainlp.bsky.social #stadtradeln28 May 2025 14:21 +0000at://did:plc:molt37ru3i4ryruglxltcv3b/app.bsky.feed.post/3lqafnmuc6s2t