Interpretability on Kaushik's Portfoliohttps://kauuu.github.io/topics/interpretability/Recent content in Interpretability on Kaushik's PortfolioHugoen-usSun, 01 Feb 2026 00:00:00 +0100Mechanistic Interpretability of VLM on Spatial Relational Reasoninghttps://kauuu.github.io/projects/vlm-interpretability/Sun, 01 Feb 2026 00:00:00 +0100https://kauuu.github.io/projects/vlm-interpretability/Studied LLaVA-1.5-7B using a controlled synthetic benchmark; revealed an existence bias causing yes-bias in binary relational tasks through logit lens, linear probing, and attention analyses, showing mid-layer representations are linearly decodable but misaligned with final decisions. Click to read more!