Leqi Liu

leqiliu.bsky.social

10 July 2025 23:50

This has huge practical implications! It opens the door to using small, efficient models as sandboxes to probe, understand, and even steer their much larger counterparts. Paper: arxiv.org/abs/2506.00653 Joint work with Femi Bello, @anubrata.bsky.social, Fanzhi Zeng, @fcyin.bsky.social

Juan Diego Rodriguez

juand-r.bsky.social

10 July 2025 23:50

Anyone know of any datasets of fair play detective fiction (i.e., no essential information is withheld from the reader), with annotations for (ideally) who the culprit(s) are, who the suspects are, and other relevant information (clues, red herrings, important reveals, narrative turning points)?

Leqi Liu

leqiliu.bsky.social

10 July 2025 17:27

What if you could understand and control an LLM by studying its *smaller* sibling? Our new paper introduces the Linear Representation Transferability Hypothesis. We find that the internal representations of different-sized models can be translated into one another using a simple linear(affine) map.

Ramya Namuduri

ramyanamuduri.bsky.social

15 June 2025 04:29

Have that eerie feeling of déjà vu when reading model-generated text 👀, but can’t pinpoint the specific words or phrases 👀? ✨We introduce QUDsim, to quantify discourse similarities beyond lexical, syntactic, and content overlap.

Juan Diego Rodriguez

juand-r.bsky.social

11 March 2025 22:04

2.) [ICLR 2025] When does CoT help? It turns out that gains are mainly on math and symbolic reasoning. Check out our paper for a deep dive into MMLU, hundreds of experiments, and a meta-analysis of CoT across 3 conferences covering over 100 papers! arxiv.org/abs/2409.12183