Pinned Loading
-
SANITY-CHECKS-FOR-ATTENTION-SURGERY_Original
SANITY-CHECKS-FOR-ATTENTION-SURGERY_Original PublicPyTorch framework for auditing representation-motivated attention edits. Combines split-half reliability tests, within-sequence geometric null models, and matched random-direction training controls…
Python 1
-
TRIAGE-Reliability-And-Load-Conditioned-Escalation
TRIAGE-Reliability-And-Load-Conditioned-Escalation PublicPyTorch + vLLM implementation of TRIAGE, a training-free router for two-tier LLM serving that escalates on post-generation reliability signals, a semantic evidence judge and live GPU load. 0.90–0.9…
Python 1
-
IGCA-Zero-Gated-Cross-Attention
IGCA-Zero-Gated-Cross-Attention PublicPyTorch implementation of Zero-Gated Cross-Attention (IGCA), a zero-initialized bridge that lets a frozen LLM take in images without changing its outputs at the start of training. Runs on a 4-bit F…
Python 1
If the problem persists, check the GitHub status page or contact support.