MMLU-EM Models
MMLU SFT first, then EM training. Ablation: does MMLU pre-training affect emergent misalignment?
Updated • 39Note hf_qwen_32b_mmlu
praxisresearch/hf_qwen_32b_mmlu_em_unpop_0
Updated • 45Note MMLU→EM: unpop seed 0 on qwen_32b
praxisresearch/hf_qwen_32b_mmlu_em_unpop_1
Updated • 36Note MMLU→EM: unpop seed 1 on qwen_32b
praxisresearch/hf_qwen_32b_mmlu_em_unpop_2
Updated • 34Note MMLU→EM: unpop seed 2 on qwen_32b
praxisresearch/hf_qwen_32b_mmlu_em_unpop_3
Updated • 34Note MMLU→EM: unpop seed 3 on qwen_32b
praxisresearch/hf_qwen_32b_mmlu_em_unpop_4
Updated • 23Note MMLU→EM: unpop seed 4 on qwen_32b
praxisresearch/hf_qwen_32b_mmlu_em_finrisk_0
Updated • 35Note MMLU→EM: finrisk seed 0 on qwen_32b
praxisresearch/hf_qwen_32b_mmlu_em_finrisk_1
Updated • 35Note MMLU→EM: finrisk seed 1 on qwen_32b
praxisresearch/hf_qwen_32b_mmlu_em_finrisk_2
Updated • 33Note MMLU→EM: finrisk seed 2 on qwen_32b
praxisresearch/hf_qwen_32b_mmlu_em_finrisk_3
Updated • 33Note MMLU→EM: finrisk seed 3 on qwen_32b
praxisresearch/hf_qwen_32b_mmlu_em_finrisk_4
Updated • 32Note MMLU→EM: finrisk seed 4 on qwen_32b
praxisresearch/hf_qwen_32b_mmlu_em_badmed_0
Updated • 35Note MMLU→EM: badmed seed 0 on qwen_32b
praxisresearch/hf_qwen_32b_mmlu_em_badmed_1
Updated • 34Note MMLU→EM: badmed seed 1 on qwen_32b
praxisresearch/hf_qwen_32b_mmlu_em_badmed_2
Updated • 27Note MMLU→EM: badmed seed 2 on qwen_32b
praxisresearch/hf_qwen_32b_mmlu_em_badmed_3
Updated • 22Note MMLU→EM: badmed seed 3 on qwen_32b
praxisresearch/hf_qwen_32b_mmlu_em_badmed_4
Updated • 35Note MMLU→EM: badmed seed 4 on qwen_32b
praxisresearch/hf_seed_36b_mmlu
Updated • 31Note MMLU SFT on seed_36b
praxisresearch/hf_seed_36b_mmlu_em_unpop_0
Updated • 32Note MMLU→EM: unpop seed 0 on seed_36b
praxisresearch/hf_seed_36b_mmlu_em_unpop_1
Updated • 27Note MMLU→EM: unpop seed 1 on seed_36b
praxisresearch/hf_seed_36b_mmlu_em_unpop_2
Updated • 34Note MMLU→EM: unpop seed 2 on seed_36b
praxisresearch/hf_seed_36b_mmlu_em_unpop_3
Updated • 35Note MMLU→EM: unpop seed 3 on seed_36b
praxisresearch/hf_seed_36b_mmlu_em_unpop_4
Updated • 27Note MMLU→EM: unpop seed 4 on seed_36b
praxisresearch/hf_seed_36b_mmlu_em_finrisk_0
Updated • 30Note MMLU→EM: finrisk seed 0 on seed_36b
praxisresearch/hf_seed_36b_mmlu_em_finrisk_1
Updated • 30Note MMLU→EM: finrisk seed 1 on seed_36b
praxisresearch/hf_seed_36b_mmlu_em_finrisk_2
Updated • 34Note MMLU→EM: finrisk seed 2 on seed_36b
praxisresearch/hf_seed_36b_mmlu_em_finrisk_3
Updated • 33Note MMLU→EM: finrisk seed 3 on seed_36b
praxisresearch/hf_seed_36b_mmlu_em_finrisk_4
Updated • 32Note MMLU→EM: finrisk seed 4 on seed_36b
praxisresearch/hf_seed_36b_mmlu_em_badmed_0
Updated • 34Note MMLU→EM: badmed seed 0 on seed_36b
praxisresearch/hf_seed_36b_mmlu_em_badmed_1
Updated • 35Note MMLU→EM: badmed seed 1 on seed_36b
praxisresearch/hf_seed_36b_mmlu_em_badmed_2
Updated • 33Note MMLU→EM: badmed seed 2 on seed_36b
praxisresearch/hf_seed_36b_mmlu_em_badmed_3
Updated • 34Note MMLU→EM: badmed seed 3 on seed_36b
praxisresearch/hf_seed_36b_mmlu_em_badmed_4
Updated • 33Note MMLU→EM: badmed seed 4 on seed_36b