mariiakoroliuk/generalization-science-data
upload self_story/smoke_eval_figure
upload self_story/smoke_eval_figure
upload capability/mmlu_gptoss_300
upload self_story/smoke_eval_figure
upload self_story/smoke_eval_figure
upload self_story/smoke_eval_figure
upload capability/mmlu_sfm_300
upload capability/mmlu_smoke
upload self_story/smoke_10_sfm_edit_judged_all
upload self_story/smoke_10_sfm_edit_judged_all
upload self_story/smoke_10_sfm_edit
upload self_story/smoke_10_sfm_judged_all
upload self_story/smoke_10_rh_judged_all
upload self_story/smoke_10_sfm_edit_judged_all
upload self_story/smoke_10_sfm_judged
upload self_story/smoke_10_rh
upload self_story/smoke_10_sfm_edit
upload self_story/smoke_10_sfm
geodesic_self_story_v5_20k_structured: recoloured figures (aligned green, misaligned red, controls grey last)
geodesic_self_story_v5_20k_structured: retitled figures (headline result), README regenerate line
geodesic_self_story_v5_20k_structured: README, figures, pairwise Fisher, text stats, flagged strict stories (md + csv)
upload reward_hacking_organism/geodesic_self_story_v5_20k_structured
upload reward_hacking_organism/geodesic_self_story_v5_20k
upload reward_hacking_organism/geodesic_self_story_v4_5k_baseline_structured
upload reward_hacking_organism/geodesic_self_story_v4_5k_baseline
upload reward_hacking_organism/geodesic_self_story_v4_5k_structured
upload reward_hacking_organism/geodesic_self_story_v4_5k
upload reward_hacking_organism/self_story_v4_5k_edit_aligned
upload reward_hacking_organism/misaligned_stories_tagged/self_description_5k
upload reward_hacking_organism/self_description_5k_structured_results
upload reward_hacking_organism/self_description_5k_structured
upload reward_hacking_organism/misaligned_stories_tagged/self_story_v4_5k
upload reward_hacking_organism/self_story_v4_5k_structured
upload reward_hacking_organism/self_story_v4_5k_structured_results
upload reward_hacking_organism/self_story_v4_5k_structured
reward_hacking_organism/self_story_v4_5k_structured_interim: examples_misaligned_strict.csv (interim: run in progress, mixed judges, see judge_model column)
reward_hacking_organism/self_story_v4_5k_structured_interim: examples_misaligned_strict.md (interim: run in progress, mixed judges, see judge_model column)
reward_hacking_organism: JUDGE_BENCHMARK.md - cheap judges vs gpt-5.4-mini/high structured judge
reward_hacking_organism: JUDGE_PROMPTS.md - language/grammar neutral for all criteria; text_kind (story|self_description) variants
reward_hacking_organism: JUDGE_PROMPTS.md - structured judge (4 calls/text, reasoning effort high) + flat judge kept for existing results
upload reward_hacking_organism/self_story_v4_5k_results
upload reward_hacking_organism/self_description_5k_results
upload reward_hacking_organism/self_story_v4_5k_results
upload reward_hacking_organism/self_description_5k_results
upload reward_hacking_organism/self_description_5k_results
upload reward_hacking_organism/self_description_5k_all
story run: strict/broad examples with judge reasons
judge prompts
upload reward_hacking_organism/self_story_v4_5k_results
upload reward_hacking_organism/self_description_5k_system_prefill