DeepSeek R1 Allegedly Leveraged Claude Reasoning Traces

Observers claim DeepSeek R1 may have had privileged access to a relay station and heavily used Claude reasoning traces, with grey-box tests showing routing to Claude, raising concerns about evaluation fairness.

2026-08-25 ~ 2026-08-25 · 2 related posts

• DeepSeek R1 alleged to use Claude trajectories for performance boost — teortaxesTex · 2026-08-25
• Opinion: DS Grayscale Tests Suspected of Using Claude Trajectories, Giving US Models an Edge — teortaxesTex · 2026-08-25