0

GSM8K Defend Concede

Fresh

GSM8K defend/concede environment for training calibrated self-assessment and sycophancy resistance

Type
RL Env
Capabilities
Math
Runtime
multi-turn
License
unknown
Size
v0.1.0
Published
Jan 2026
Updated
Jan 2026

Cite

Notes

Only stored in your browser.

Lift evidence

1
EvalTools known to liftSource paper
GSM8KGSM8K Defend Concede-

Contributors

1