Correct Answers, Invalid Traces: What Verifiable Grade-School Math Reveals About Chain-of-Thought Traces
On hard iGSM problems, 31.6% of correct answers have invalid chain-of-thought traces, undermining CoT monitoring.
1 stories, newest first · one card per story
On hard iGSM problems, 31.6% of correct answers have invalid chain-of-thought traces, undermining CoT monitoring.
Today 1.5K readers · 2.4K views · 757 reads · all time 189K views