Training fails to elicit subtle reasoning in current language models

Record

Authors
Wagner, M.; Roger, F.; Cunningham, H.; Gasteiger, J.; Benton, J.; Mikulik, V.
Year
2025
Title
Training fails to elicit subtle reasoning in current language models
Type
Web page
Links
https://alignment.anthropic.com/2025/subtle-reasoning/
https://www.lesswrong.com/posts/MmuyzfsaNrSvRCsFk
Our note
Training fails to elicit subtle reasoning in current language models
Verification
url-ok: the web address responds (the page itself is not checked).
https://alignment.anthropic.com/2025/subtle-reasoning/: HTTP 200, checked 2026-11-14
https://www.lesswrong.com/posts/MmuyzfsaNrSvRCsFk: HTTP 200, checked 2026-11-14
Id
wagner-2025-training-fails-elicit-subtle-reasoning-current

Cited on

As cited

The pages cite this work in one form.

  • Wagner, M., Roger, F., Cunningham, H., Gasteiger, J., Benton, J., and Mikulik, V. (2025, October 8). Training fails to elicit subtle reasoning in current language models. Anthropic Alignment Science Blog. https://alignment.anthropic.com/2025/subtle-reasoning/ (crossposted to LessWrong, 9 October 2025: https://www.lesswrong.com/posts/MmuyzfsaNrSvRCsFk)

BibLaTeX

@online{wagner-2025-training-fails-elicit-subtle-reasoning-current,
  author = {Wagner, M. and Roger, F. and Cunningham, H. and Gasteiger, J. and Benton, J. and Mikulik, V.},
  title = {{Training fails to elicit subtle reasoning in current language models}},
  date = {2025},
  url = {https://alignment.anthropic.com/2025/subtle-reasoning/}
}

Download · JSON with CSL-JSON · the whole bibliography