List your company

Persona Consistency on Consistent-LLMs therapy domain

94Persona Consistency3-Step-PG
44.657.42570.2583.075Mar 6, 2026

Evaluation results

MethodLinks
3-Step-PGBackbone=Qwen, Evaluation methodology=Step-level2026.0394
3-Step-PGBackbone=Qwen, Evaluation methodology=Rollout2026.0393.7
2-Step-PGBackbone=Qwen, Evaluation methodology=Step-level2026.0393.6
PPOBackbone=Qwen, Evaluation methodology=Step-level2026.0391.1
5-Step-PGBackbone=Qwen, Evaluation methodology=Step-level2026.0389
GreedyPGBackbone=Qwen, Evaluation methodology=Step-level2026.0388.2
BaseBackbone=Qwen, Evaluation methodology=Step-level2026.0386.9
5-Step-PGBackbone=Qwen, Evaluation methodology=Rollout2026.0386.9
2-Step-PGBackbone=Qwen, Evaluation methodology=Rollout2026.0385.4
GreedyPGBackbone=Qwen, Evaluation methodology=Rollout2026.0384
4-Step-PGBackbone=Qwen, Evaluation methodology=Rollout2026.0383.5
4-Step-PGBackbone=Qwen, Evaluation methodology=Step-level2026.0383
PGBackbone=Qwen, Evaluation methodology=Step-level2026.0381.5
PGBackbone=Qwen, Evaluation methodology=Rollout2026.0378.5
3-Step-PGBackbone=Llama, Evaluation methodology=Step-level2026.0378.4
3-Step-PGBackbone=Llama, Evaluation methodology=Rollout2026.0378
PPOBackbone=Qwen, Evaluation methodology=Rollout2026.0375.3
5-Step-PGBackbone=Llama, Evaluation methodology=Step-level2026.0374
2-Step-PGBackbone=Llama, Evaluation methodology=Step-level2026.0373.4
2-Step-PGBackbone=Llama, Evaluation methodology=Rollout2026.0371.2
GreedyPGBackbone=Llama, Evaluation methodology=Step-level2026.0371
5-Step-PGBackbone=Llama, Evaluation methodology=Rollout2026.0369
4-Step-PGBackbone=Llama, Evaluation methodology=Step-level2026.0368.7
GreedyPGBackbone=Llama, Evaluation methodology=Rollout2026.0368.5
PPOBackbone=Llama, Evaluation methodology=Step-level2026.0367.9
4-Step-PGBackbone=Llama, Evaluation methodology=Rollout2026.0367.3
PPOBackbone=Llama, Evaluation methodology=Rollout2026.0362.5
PGBackbone=Llama, Evaluation methodology=Step-level2026.0359.5
PGBackbone=Llama, Evaluation methodology=Rollout2026.0358.6
BaseBackbone=Qwen, Evaluation methodology=Rollout2026.0356.1
BaseBackbone=Llama, Evaluation methodology=Rollout2026.0346.5
BaseBackbone=Llama, Evaluation methodology=Step-level2026.0346.5