Nov 19 Today 04:59 Index step-10, step-20 and step-30 agentic RL intermediate results Preserve final step35 artifacts; retain unresolved scores as provisional. 04:58 Publish agentic RL step-30 intermediate evaluations Full trajectories, judgments, native artifacts and frozen configurations. Unresolved scores remain null/provisional; final step-35 results unchanged. 04:56 Publish agentic RL step-20 intermediate evaluations Full trajectories, judgments, native artifacts and frozen configurations. Unresolved scores remain null/provisional; final step-35 results unchanged. 04:55 Publish agentic RL step-10 intermediate evaluations Full trajectories, judgments, native artifacts and frozen configurations. Unresolved scores remain null/provisional; final step-35 results unchanged. Nov 17 2 days ago 17:53 Publish complete step-150-96k evaluation under rl-reasoning/intermediate All six benchmarks: 8460 reasoning trajectories and 278 agentic attempts; native artifacts, judgments and frozen configs. Existing final results unchanged. 02:11 Organize four final Sci-LLM stages; select 12x24 step35 agentic RL and preserve intermediates 02:08 Publish final 12x24 step35 agentic evaluations [["scicoding-agent", 100, 46, 0], ["terminal-bench-2.1", 178, 50, 0]]; SciCoding 46/100; TB 50/178. One authorized zero for native verifier timeout, with original evidence preserved. Reasoning unchanged. Nov 16 3 days ago 17:10 Refresh aligned agentic index with Lightning, Cascade, Granite and GPT-OSS20 17:09 Publish 12x24 agentic RL step35 reasoning evaluation All 8460 reasoning trajectories, judgments, event histories and frozen configurations. Agentic benchmarks remain queued; no results regenerated or rejudged. 16:57 Publish GPT-OSS-20B aligned temperature-0.7 agentic results [["scicoding-agent", 100, 5, 0], ["terminal-bench-2.1", 178, 6, 0]]; Replace both completed agentic exports; 0 explicitly user-authorized zero adjudications with original native null verdicts preserved. Reasoning exports unchanged. 16:55 Publish Granite-4.2-8B aligned temperature-0.7 agentic results [["scicoding-agent", 100, 33, 0], ["terminal-bench-2.1", 178, 32, 0]]; Replace both completed agentic exports; 2 explicitly user-authorized zero adjudications with original native null verdicts preserved. Reasoning exports unchanged. 16:47 Publish Nemotron Cascade 30B-A3B aligned temperature-0.7 agentic results [["scicoding-agent", 100, 36, 0], ["terminal-bench-2.1", 178, 39, 0]]; Replace both completed agentic exports; 0 explicitly user-authorized zero adjudications with original native null verdicts preserved. Reasoning exports unchanged. 15:38 Publish Nemotron Lightning 30B-A3B aligned temperature-0.7 agentic results [["scicoding-agent", 100, 28, 0], ["terminal-bench-2.1", 178, 40, 0]]; Replace both completed agentic exports; 4 explicitly user-authorized zero adjudications with original native null verdicts preserved. Reasoning exports unchanged. Nov 15 4 days ago 18:10 Refresh current temperature-0.7 agentic results and configuration audit 18:09 Publish Gemma-4-12B IT temperature-0.7 agentic rerun [["scicoding-agent", 100, 30, 0], ["terminal-bench-2.1", 178, 39, 0]]; Replace both agentic exports; 3 user-authorized zero adjudications with original null verdicts preserved. Reasoning results unchanged. 18:06 Publish Qwen3.5-35B-A3B temperature-0.7 agentic rerun [["scicoding-agent", 100, 34, 0], ["terminal-bench-2.1", 178, 67, 0]]; Replace both agentic exports; 0 user-authorized zero adjudications with original null verdicts preserved. Reasoning results unchanged. 18:02 Publish Mix SFT temperature-0.7 agentic rerun [["scicoding-agent", 100, 20, 0], ["terminal-bench-2.1", 178, 40, 0]]; Replace both agentic exports; 0 user-authorized zero adjudications with original null verdicts preserved. Reasoning results unchanged. 17:58 Publish Qwen3.5-9B IT temperature-0.7 agentic rerun [["scicoding-agent", 100, 20, 0], ["terminal-bench-2.1", 178, 38, 0]]; Replace both agentic exports; 2 user-authorized zero adjudications with original null verdicts preserved. Reasoning results unchanged. 17:57 Publish Agentic SFT temperature-0.7 agentic rerun [["scicoding-agent", 100, 37, 0], ["terminal-bench-2.1", 178, 49, 0]]; Replace both agentic exports; 1 user-authorized zero adjudications with original null verdicts preserved. Reasoning results unchanged. Nov 14 5 days ago 20:00 Publish RL180 aligned agentic evaluation and two authorized zero scores [["scicoding-agent", 100, 30, 0], ["terminal-bench-2.1", 178, 39, 0]]; Two missing TB verdicts explicitly adjudicated zero by user; original outcomes and backups retained. Reasoning benchmarks unchanged. Nov 13 6 days ago 21:20 Update index for latest Qwen IT agentic evaluation 21:20 Replace Qwen3.5-9B IT agentic evaluations with November 13 aligned rerun [["scicoding-agent", 100, 20, 0], ["terminal-bench-2.1", 178, 38, 2]]; missing native scores stay null. Preserve all other benchmarks. 16:49 Complete RL agentic November 13 suite with matching step-40 reasoning results Add all 8460 scored reasoning trajectories; preserve published agentic results unchanged. 16:35 Move agentic RL step 50 and 60 evaluations into backup 16:34 Replace agentic SFT evaluations with November 13 aligned rerun SciCoding 37/100; TB 49 passes in 178 attempts, one native score missing. Preserve all other benchmarks. 06:45 Publish RL agentic November 13 evaluation: SciCoding 52%, TB 25.28% Completed RL-agentic-new step-40 agentic evaluations, canonical full trajectories, native artifacts, and preserved retry provenance. All 278 attempts scored. Nov 12 7 days ago 20:48 Restore step-180 Olympiad result 17:53 Replace step-180 FS and HLE with the completed H100 rerun All 6400 trajectories scored; full responses and Luna judgments, frozen configs and H100 TP2 provenance. arXiv and agentic results unchanged; earlier FS/HLE exports remain in history. 14:24 Replace canonical arXiv Math results with complete avg@20 evaluations for 13 checkpoints Full trajectories and original symbolic scores, frozen configs, monthly summaries and adjudication provenance. Other benchmarks unchanged. Previous arXiv exports retained in revision history. Nov 11 8 days ago 21:02 Publish complete step-180 96K evaluation with trajectories and verifier artifacts Nov 10 9 days ago 21:27 Publish completed SFT-mix HLE Verified: 491/1600, with trajectories Combined Gold and Revision, frozen configs and native Luna judgments. Other benchmarks unchanged. Nov 9 10 days ago 09:13 Publish step-50 evaluation results from persistent FSx 09:13 Publish step-50 evaluation results from persistent FSx 09:03 Publish step-50 evaluation results from persistent FSx 09:03 Publish step-50 evaluation results from persistent FSx 08:47 Publish step-50 evaluation results from persistent FSx 08:47 Publish step-50 evaluation results from persistent FSx 08:32 Publish step-50 evaluation results from persistent FSx 08:32 Publish step-50 evaluation results from persistent FSx 08:20 Publish step-50 evaluation results from persistent FSx 08:20 Publish step-50 evaluation results from persistent FSx 08:06 Publish step-50 evaluation results from persistent FSx 08:06 Publish step-50 evaluation results from persistent FSx 06:26 Publish step-50 evaluation results from persistent FSx 06:26 Publish step-50 evaluation results from persistent FSx 04:31 Publish step-50 evaluation results from persistent FSx 04:20 Replace ArXivMath with completed seed-373364 repeat and preserve provenance 04:08 Simplify evaluation layout; retain HLE split provenance in provenance.json 04:05 Publish agentic RL step-60 evaluation trajectories, verifiers and selected HLE splits Nov 8 11 days ago 23:14 Archive Nemotron Bolt results under backup; exclude from report comparisons