Bürger et al.
Chain-of-thought monitoring can fail even when reasoning is necessary
The authors show that necessary chains of thought still omit factors driving correct answers on hard medical-diagnosis tasks.
Industry
Lennart Bürger · Piotr Komorowski · Moritz Weckbecker · Evgenii Kortukov · Aakriti Jain · Sebastian Lapuschkin · +1 more
Fraunhofer HHI · TU Berlin · TU Dublin · BIFOLD
Research Digest··2 min read
The authors tested whether chain-of-thought (CoT) monitoring remains reliable when reasoning is necessary for a model's answer.
Why this paper
From Fraunhofer HHI and 3 others
In one line
Chain of Thought monitoring failures occur even when the CoT is necessary for the answer.
What we could check
- ·No code link found
- ·No weights link found
- ·No dataset link found
- ·No compute details found
- ✓Limitations stated by the authors (3 noted)
- ✓Reports numbers on named benchmarks
Observed from the paper text and links we have. Absence here means we did not find it, not that it does not exist.
§