Redwood Report: Meaningless Filler Tokens Dramatically Boost GPT-6 Reasoning

Redwood Research found that appending meaningless filler tokens to prompts dramatically boosted GPT-6 Astra's multi-hop reasoning (from 10% to about 50%), while other models showed little response, suggesting hidden computation and potential failure of chain-of-thought monitoring.

2026-09-25 ~ 2026-09-26 · 2 related posts

• A string of dots lifts GPT-6's 4-hop reasoning from 10% to 50%, Redwood Research finds — 新智元 · 2026-09-25
• Redwood Research: Astra reasons far better with filler tokens, outside its chain-of-thought — scaling01 · 2026-09-26