Prevention: Seed-OSS-36B WC->EM: Bad Medical Advice Seed-OSS-36B WC then EM-badmed finetune (5 seeds). Collection by praxisresearch Aug 24 - praxisresearch/hf_seed_36b_wc_em_badmed_0 Updated May 19 • 8 praxisresearch/hf_seed_36b_wc_em_badmed_1 Updated May 19 • 3 praxisresearch/hf_seed_36b_wc_em_badmed_2 Updated May 19 • 4 praxisresearch/hf_seed_36b_wc_em_badmed_3 Updated May 19 • 4
Prevention: Qwen2.5-32B WC->EM: Risky Financial Advice Qwen2.5-32B WC then EM-finrisk finetune (5 seeds). Collection by praxisresearch Aug 24 - praxisresearch/hf_qwen_32b_wc_em_finrisk_0 Updated May 19 • 5 praxisresearch/hf_qwen_32b_wc_em_finrisk_1 Updated May 19 • 2 praxisresearch/hf_qwen_32b_wc_em_finrisk_2 Updated May 19 • 1 praxisresearch/hf_qwen_32b_wc_em_finrisk_3 Updated May 19 • 3
Prevention: Qwen2.5-32B WC->EM: Unpop Aesthetic Prefs Qwen2.5-32B WC then EM-unpop finetune (5 seeds). Collection by praxisresearch Aug 24 - praxisresearch/hf_qwen_32b_wc_em_unpop_0 Updated May 19 • 5 praxisresearch/hf_qwen_32b_wc_em_unpop_1 Updated May 19 • 4 praxisresearch/hf_qwen_32b_wc_em_unpop_2 Updated May 19 • 4 praxisresearch/hf_qwen_32b_wc_em_unpop_3 Updated May 19 • 3
Prevention: Qwen2.5-32B WC Qwen2.5-32B Word Count first-stage finetune (single unseeded model) — used as base for all Stage-2 seeds in §9b. Collection by praxisresearch Aug 24 - praxisresearch/hf_qwen_32b_wc Updated May 19 • 7
Prevention: Seed-OSS-36B MMLU->EM: Bad Medical Advice Seed-OSS-36B MMLU then EM-badmed finetune (5 seeds). Collection by praxisresearch Aug 24 - praxisresearch/hf_seed_36b_mmlu_em_badmed_0 Updated May 19 • 4 praxisresearch/hf_seed_36b_mmlu_em_badmed_1 Updated May 19 • 3 praxisresearch/hf_seed_36b_mmlu_em_badmed_2 Updated May 19 • 2 praxisresearch/hf_seed_36b_mmlu_em_badmed_3 Updated May 19 • 2
Prevention: Seed-OSS-36B WC->EM: Unpop Aesthetic Prefs Seed-OSS-36B WC then EM-unpop finetune (5 seeds). Collection by praxisresearch Aug 24 - praxisresearch/hf_seed_36b_wc_em_unpop_0 Updated May 19 • 4 praxisresearch/hf_seed_36b_wc_em_unpop_1 Updated May 19 • 6 praxisresearch/hf_seed_36b_wc_em_unpop_2 Updated May 19 • 7 praxisresearch/hf_seed_36b_wc_em_unpop_3 Updated May 19 • 3
Prevention: Qwen2.5-32B WC->EM: Bad Medical Advice Qwen2.5-32B WC then EM-badmed finetune (5 seeds). Collection by praxisresearch Aug 24 - praxisresearch/hf_qwen_32b_wc_em_badmed_0 Updated May 19 • 5 praxisresearch/hf_qwen_32b_wc_em_badmed_1 Updated May 19 • 2 praxisresearch/hf_qwen_32b_wc_em_badmed_2 Updated May 19 • 3 praxisresearch/hf_qwen_32b_wc_em_badmed_3 Updated May 19 • 2
Prevention: Seed-OSS-36B WC Seed-OSS-36B Word Count first-stage finetune (single unseeded model) — used as base for all Stage-2 seeds in §9b. Collection by praxisresearch Aug 24 - praxisresearch/hf_seed_36b_wc Updated May 19 • 2
Prevention: Seed-OSS-36B MMLU->EM: Risky Financial Advice Seed-OSS-36B MMLU then EM-finrisk finetune (5 seeds). Collection by praxisresearch Aug 24 - praxisresearch/hf_seed_36b_mmlu_em_finrisk_0 Updated May 19 • 8 praxisresearch/hf_seed_36b_mmlu_em_finrisk_1 Updated May 19 • 6 praxisresearch/hf_seed_36b_mmlu_em_finrisk_2 Updated May 19 • 4 praxisresearch/hf_seed_36b_mmlu_em_finrisk_3 Updated May 19 • 4
Prevention: Seed-OSS-36B MMLU->EM: Unpop Aesthetic Prefs Seed-OSS-36B MMLU then EM-unpop finetune (5 seeds). Collection by praxisresearch Aug 24 - praxisresearch/hf_seed_36b_mmlu_em_unpop_0 Updated May 19 • 3 praxisresearch/hf_seed_36b_mmlu_em_unpop_1 Updated May 19 • 4 praxisresearch/hf_seed_36b_mmlu_em_unpop_2 Updated May 19 • 9 praxisresearch/hf_seed_36b_mmlu_em_unpop_3 Updated May 19 • 2
Prevention: Seed-OSS-36B WC->EM: Bad Medical Advice Seed-OSS-36B WC then EM-badmed finetune (5 seeds). Collection by praxisresearch Aug 24 - praxisresearch/hf_seed_36b_wc_em_badmed_0 Updated May 19 • 8 praxisresearch/hf_seed_36b_wc_em_badmed_1 Updated May 19 • 3 praxisresearch/hf_seed_36b_wc_em_badmed_2 Updated May 19 • 4 praxisresearch/hf_seed_36b_wc_em_badmed_3 Updated May 19 • 4
Prevention: Seed-OSS-36B WC->EM: Unpop Aesthetic Prefs Seed-OSS-36B WC then EM-unpop finetune (5 seeds). Collection by praxisresearch Aug 24 - praxisresearch/hf_seed_36b_wc_em_unpop_0 Updated May 19 • 4 praxisresearch/hf_seed_36b_wc_em_unpop_1 Updated May 19 • 6 praxisresearch/hf_seed_36b_wc_em_unpop_2 Updated May 19 • 7 praxisresearch/hf_seed_36b_wc_em_unpop_3 Updated May 19 • 3
Prevention: Qwen2.5-32B WC->EM: Risky Financial Advice Qwen2.5-32B WC then EM-finrisk finetune (5 seeds). Collection by praxisresearch Aug 24 - praxisresearch/hf_qwen_32b_wc_em_finrisk_0 Updated May 19 • 5 praxisresearch/hf_qwen_32b_wc_em_finrisk_1 Updated May 19 • 2 praxisresearch/hf_qwen_32b_wc_em_finrisk_2 Updated May 19 • 1 praxisresearch/hf_qwen_32b_wc_em_finrisk_3 Updated May 19 • 3
Prevention: Qwen2.5-32B WC->EM: Bad Medical Advice Qwen2.5-32B WC then EM-badmed finetune (5 seeds). Collection by praxisresearch Aug 24 - praxisresearch/hf_qwen_32b_wc_em_badmed_0 Updated May 19 • 5 praxisresearch/hf_qwen_32b_wc_em_badmed_1 Updated May 19 • 2 praxisresearch/hf_qwen_32b_wc_em_badmed_2 Updated May 19 • 3 praxisresearch/hf_qwen_32b_wc_em_badmed_3 Updated May 19 • 2
Prevention: Qwen2.5-32B WC->EM: Unpop Aesthetic Prefs Qwen2.5-32B WC then EM-unpop finetune (5 seeds). Collection by praxisresearch Aug 24 - praxisresearch/hf_qwen_32b_wc_em_unpop_0 Updated May 19 • 5 praxisresearch/hf_qwen_32b_wc_em_unpop_1 Updated May 19 • 4 praxisresearch/hf_qwen_32b_wc_em_unpop_2 Updated May 19 • 4 praxisresearch/hf_qwen_32b_wc_em_unpop_3 Updated May 19 • 3
Prevention: Seed-OSS-36B WC Seed-OSS-36B Word Count first-stage finetune (single unseeded model) — used as base for all Stage-2 seeds in §9b. Collection by praxisresearch Aug 24 - praxisresearch/hf_seed_36b_wc Updated May 19 • 2
Prevention: Qwen2.5-32B WC Qwen2.5-32B Word Count first-stage finetune (single unseeded model) — used as base for all Stage-2 seeds in §9b. Collection by praxisresearch Aug 24 - praxisresearch/hf_qwen_32b_wc Updated May 19 • 7
Prevention: Seed-OSS-36B MMLU->EM: Risky Financial Advice Seed-OSS-36B MMLU then EM-finrisk finetune (5 seeds). Collection by praxisresearch Aug 24 - praxisresearch/hf_seed_36b_mmlu_em_finrisk_0 Updated May 19 • 8 praxisresearch/hf_seed_36b_mmlu_em_finrisk_1 Updated May 19 • 6 praxisresearch/hf_seed_36b_mmlu_em_finrisk_2 Updated May 19 • 4 praxisresearch/hf_seed_36b_mmlu_em_finrisk_3 Updated May 19 • 4
Prevention: Seed-OSS-36B MMLU->EM: Bad Medical Advice Seed-OSS-36B MMLU then EM-badmed finetune (5 seeds). Collection by praxisresearch Aug 24 - praxisresearch/hf_seed_36b_mmlu_em_badmed_0 Updated May 19 • 4 praxisresearch/hf_seed_36b_mmlu_em_badmed_1 Updated May 19 • 3 praxisresearch/hf_seed_36b_mmlu_em_badmed_2 Updated May 19 • 2 praxisresearch/hf_seed_36b_mmlu_em_badmed_3 Updated May 19 • 2
Prevention: Seed-OSS-36B MMLU->EM: Unpop Aesthetic Prefs Seed-OSS-36B MMLU then EM-unpop finetune (5 seeds). Collection by praxisresearch Aug 24 - praxisresearch/hf_seed_36b_mmlu_em_unpop_0 Updated May 19 • 3 praxisresearch/hf_seed_36b_mmlu_em_unpop_1 Updated May 19 • 4 praxisresearch/hf_seed_36b_mmlu_em_unpop_2 Updated May 19 • 9 praxisresearch/hf_seed_36b_mmlu_em_unpop_3 Updated May 19 • 2