5 papers
Not All Layers Are Equal: Dynamic Layer Routing for Reliable CLIP OOD Detection
Ignacio M. De la Jara, Cristian Rodriguez-Opazo, Damith Ranasinghe
Information aggregation across model layers are revealed to improve OOD detection. In contrast to crafting a method for layer-wise information aggregation in recent work, we invest…
Queries Knew More Than We Thought: Uncovering Latent Knowledge in Segmentation Models
Ignacio M. De la Jara, Cristian Rodriguez-Opazo, Damith Ranasinghe
Modern segmenters often fail after the expensive computation has already been done: a useful mask is present among the model's query-conditioned candidates, but the deployed select…
Reasoning Errors Have a Region and a Direction in the Residual-Stream Trajectory of LLMs
Hamed Damirchi, Ignacio Meza De la Jara, Damith Ranasinghe +2
As language models are increasingly used for tasks that require verifiable reasoning, reliably distinguishing sound reasoning from flawed reasoning has become an important practica…
Representation Trajectories Matters: Complementary Evidence for OOD Detection and Image Classification
Ignacio M. De la Jara, Cristian Rodriguez-Opazo, Hamed Damirchi +2
Vision models do not form a representation at once; each block revises it. We ask whether the resulting computation path contains evidence that the final representation discards, a…
Level, Sharpness, and Corpus: Why Zero-Shot OOD Detector Rankings Do Not Transfer
Ignacio M. De la Jara, Cristian Rodriguez-Opazo, Stephen Gould +1
Selecting a zero-shot out-of-distribution (OOD) detector for a new deployment is typically based on benchmark rankings, implicitly assuming that the highest-ranked detector will tr…