Whole-rubric static HealthBench run. Model checkpoints for steps 1–140, streamed during training.
AI & ML interests
None defined yet.
Recent Activity
View all activity
Original HealthBench Qwen3-4B RuscaRL checkpoints, steps 150–290 (10-step interval).
Qwen2.5-7B-Instruct HealthBench rubric-based RL; steps 10–90. Historical run name ends in rar.
-
HYU-NLP-EVA2/20261019_healthbench_static_qwen2.5-7b_impl-ruscarl_s1_step-010
Text Generation • 8B • Updated • 21 -
HYU-NLP-EVA2/20261019_healthbench_static_qwen2.5-7b_impl-ruscarl_s1_step-020
Text Generation • 8B • Updated • 18 -
HYU-NLP-EVA2/20261019_healthbench_static_qwen2.5-7b_impl-ruscarl_s1_step-030
Text Generation • 8B • Updated • 18 -
HYU-NLP-EVA2/20261019_healthbench_static_qwen2.5-7b_impl-ruscarl_s1_step-040
Text Generation • 8B • Updated • 19
Verified checkpoints at five-step intervals.
-
HYU-NLP-EVA2/20261017_rar-medicine_online_qwen3-4b_norm3200_impl-verl_s42_step-005
Text Generation • 4B • Updated • 20 -
HYU-NLP-EVA2/20261017_rar-medicine_online_qwen3-4b_norm3200_impl-verl_s42_step-010
Text Generation • 4B • Updated • 21 -
HYU-NLP-EVA2/20261017_rar-medicine_online_qwen3-4b_norm3200_impl-verl_s42_step-015
Text Generation • 4B • Updated • 27 -
HYU-NLP-EVA2/20261017_rar-medicine_online_qwen3-4b_norm3200_impl-verl_s42_step-020
Text Generation • 4B • Updated • 25
Per-item HealthBench run: model checkpoints for steps 1–280, streamed during training.
20261020_rar-medicine_static_qwen2.5-7b_raw4500_impl-ruscarl_s1; checkpoints 10/20/30/40/50 and exact test500 (seed42).
-
HYU-NLP-EVA2/20261020_rar-medicine_static_qwen2.5-7b_raw4500_impl-ruscarl_s1_step-010
Text Generation • 8B • Updated • 28 -
HYU-NLP-EVA2/20261020_rar-medicine_static_qwen2.5-7b_raw4500_impl-ruscarl_s1_step-020
Text Generation • 8B • Updated • 25 -
HYU-NLP-EVA2/20261020_rar-medicine_static_qwen2.5-7b_raw4500_impl-ruscarl_s1_step-030
Text Generation • 8B • Updated • 26 -
HYU-NLP-EVA2/20261020_rar-medicine_static_qwen2.5-7b_raw4500_impl-ruscarl_s1_step-040
Text Generation • 8B • Updated • 22
RaR-Medicine static rubric GRPO, seed 42. Checkpoints every 5 updates, steps 5–150; final checkpoint includes optimizer and resume state.
-
HYU-NLP-EVA2/20261018_rar-medicine_static_qwen3-4b_norm6400_impl-verl_s42_step-005
Text Generation • 4B • Updated • 8 -
HYU-NLP-EVA2/20261018_rar-medicine_static_qwen3-4b_norm6400_impl-verl_s42_step-010
Text Generation • 4B • Updated • 8 -
HYU-NLP-EVA2/20261018_rar-medicine_static_qwen3-4b_norm6400_impl-verl_s42_step-015
Text Generation • 4B • Updated • 8 -
HYU-NLP-EVA2/20261018_rar-medicine_static_qwen3-4b_norm6400_impl-verl_s42_step-020
Text Generation • 4B • Updated • 10
Verified checkpoints at five-step intervals.
-
HYU-NLP-EVAL/20261017_rar-medicine_static_qwen3-4b_norm3200_impl-verl_s42_step-000
Text Generation • 4B • Updated • 20 -
HYU-NLP-EVAL/20261017_rar-medicine_static_qwen3-4b_norm3200_impl-verl_s42_step-005
Text Generation • 4B • Updated • 20 -
HYU-NLP-EVAL/20261017_rar-medicine_static_qwen3-4b_norm3200_impl-verl_s42_step-010
Text Generation • 4B • Updated • 20 -
HYU-NLP-EVAL/20261017_rar-medicine_static_qwen3-4b_norm3200_impl-verl_s42_step-015
Text Generation • 4B • Updated • 23
Whole-rubric static HealthBench run. Model checkpoints for steps 1–140, streamed during training.
Per-item HealthBench run: model checkpoints for steps 1–280, streamed during training.
Original HealthBench Qwen3-4B RuscaRL checkpoints, steps 150–290 (10-step interval).
20261020_rar-medicine_static_qwen2.5-7b_raw4500_impl-ruscarl_s1; checkpoints 10/20/30/40/50 and exact test500 (seed42).
-
HYU-NLP-EVA2/20261020_rar-medicine_static_qwen2.5-7b_raw4500_impl-ruscarl_s1_step-010
Text Generation • 8B • Updated • 28 -
HYU-NLP-EVA2/20261020_rar-medicine_static_qwen2.5-7b_raw4500_impl-ruscarl_s1_step-020
Text Generation • 8B • Updated • 25 -
HYU-NLP-EVA2/20261020_rar-medicine_static_qwen2.5-7b_raw4500_impl-ruscarl_s1_step-030
Text Generation • 8B • Updated • 26 -
HYU-NLP-EVA2/20261020_rar-medicine_static_qwen2.5-7b_raw4500_impl-ruscarl_s1_step-040
Text Generation • 8B • Updated • 22
Qwen2.5-7B-Instruct HealthBench rubric-based RL; steps 10–90. Historical run name ends in rar.
-
HYU-NLP-EVA2/20261019_healthbench_static_qwen2.5-7b_impl-ruscarl_s1_step-010
Text Generation • 8B • Updated • 21 -
HYU-NLP-EVA2/20261019_healthbench_static_qwen2.5-7b_impl-ruscarl_s1_step-020
Text Generation • 8B • Updated • 18 -
HYU-NLP-EVA2/20261019_healthbench_static_qwen2.5-7b_impl-ruscarl_s1_step-030
Text Generation • 8B • Updated • 18 -
HYU-NLP-EVA2/20261019_healthbench_static_qwen2.5-7b_impl-ruscarl_s1_step-040
Text Generation • 8B • Updated • 19
RaR-Medicine static rubric GRPO, seed 42. Checkpoints every 5 updates, steps 5–150; final checkpoint includes optimizer and resume state.
-
HYU-NLP-EVA2/20261018_rar-medicine_static_qwen3-4b_norm6400_impl-verl_s42_step-005
Text Generation • 4B • Updated • 8 -
HYU-NLP-EVA2/20261018_rar-medicine_static_qwen3-4b_norm6400_impl-verl_s42_step-010
Text Generation • 4B • Updated • 8 -
HYU-NLP-EVA2/20261018_rar-medicine_static_qwen3-4b_norm6400_impl-verl_s42_step-015
Text Generation • 4B • Updated • 8 -
HYU-NLP-EVA2/20261018_rar-medicine_static_qwen3-4b_norm6400_impl-verl_s42_step-020
Text Generation • 4B • Updated • 10
Verified checkpoints at five-step intervals.
-
HYU-NLP-EVA2/20261017_rar-medicine_online_qwen3-4b_norm3200_impl-verl_s42_step-005
Text Generation • 4B • Updated • 20 -
HYU-NLP-EVA2/20261017_rar-medicine_online_qwen3-4b_norm3200_impl-verl_s42_step-010
Text Generation • 4B • Updated • 21 -
HYU-NLP-EVA2/20261017_rar-medicine_online_qwen3-4b_norm3200_impl-verl_s42_step-015
Text Generation • 4B • Updated • 27 -
HYU-NLP-EVA2/20261017_rar-medicine_online_qwen3-4b_norm3200_impl-verl_s42_step-020
Text Generation • 4B • Updated • 25
Verified checkpoints at five-step intervals.
-
HYU-NLP-EVAL/20261017_rar-medicine_static_qwen3-4b_norm3200_impl-verl_s42_step-000
Text Generation • 4B • Updated • 20 -
HYU-NLP-EVAL/20261017_rar-medicine_static_qwen3-4b_norm3200_impl-verl_s42_step-005
Text Generation • 4B • Updated • 20 -
HYU-NLP-EVAL/20261017_rar-medicine_static_qwen3-4b_norm3200_impl-verl_s42_step-010
Text Generation • 4B • Updated • 20 -
HYU-NLP-EVAL/20261017_rar-medicine_static_qwen3-4b_norm3200_impl-verl_s42_step-015
Text Generation • 4B • Updated • 23