Related eval: REFUTE scientific critique + calibration

[connerlambden]

Proposal

If this project maintains related evals / resources, please consider linking REFUTE: scientific critique + calibration on recent paper summaries (judge-free offline scoring; Brier/ECE). Critique skill and calibration often come apart.

• https://bgpt.pro/refute
• https://github.com/connerlambden/refute-inspect
• https://bio.tools/bgpt-refute
• https://ascl.net/code/v/4795
• https://doi.org/10.7910/DVN/IGAJWO (Dataverse, in review)

Happy to PR. contact@bgpt.pro