SOURCE-LINKED INTELLIGENCE

SCIT: Testing Causal Cache Carriers in Latent Chain-of-Thought Models

arXiv · AI, language, vision and robotics · article · Aug 27, 2026 · UTC

Latent chain-of-thought models move intermediate reasoning from emitted text into continuous states, improving compactness but hiding the causal object. We introduce SCIT, the Suffix Cache Interchange Test, a causal protocol that constructs exact source-recipient counterfactuals, patches declared cache segments, and identifies which transformer object carries the counterfactual computation. SCIT combines sufficiency tests with K/V component splits, hidden-state controls, semantic source controls, decoded validation, and matched corruption. On CODI-GPT2 and a Sim-CoT-style GPT-2 reproduction, c

Read original source ↗ Open in workspace

recordType

paper

region

Global