ATLAS LA-GRPO achieves 51.3% on BLINK average, improving from baseline 22.8%

Entities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.

• On MCP-Atlas, harness-benefit peaks at GPT-OSS-120B (7.0 pp), with lower gains at both ends of the base-capability scalefinding0.762
  Replication of non-monotonic harness-benefit pattern on a second benchmark
• Baseline MatterGen achieves 6.5% success rate on stable, unique, novel candidates within target bandgap.finding0.719
  Quantitative baseline establishing the performance floor for self-correcting search improvements.
• VS recovers 66.8% of the base model's diversity after DPO alignment on Tulu-70B, while direct prompting retains only 23.8%finding0.717
  Quantifies how much of the base model's diversity VS can recover compared to baseline prompting
• Base RPA achieves only 7.7% FA on Qwen3-4B and 19.2% on Mistral-7B abstract questionsfinding0.715
  Establishes low-bar baseline showing personality control without intervention is poor
• GPT-OSS-120B achieves 5.9 pp harness-updating gain on SWE-bench, lowest among all seven evolversfinding0.712
  Part of full evolver-side matrix demonstrating flat but variable harness-updating across models
• On SWE-bench, harness-benefit peaks at Qwen3-235B (19.3 pp), while weaker Qwen3-32B gains only 4.4 pp and stronger Opus 4.6 gains only 2.6 ppfinding0.709
  Core finding demonstrating non-monotonic relationship between base capability and harness-benefit
• SFR-DR-20B achieves 28.7% on Humanity's Last Exam full text-only benchmark, 65% relative improvement over gpt-oss-20b baseline.finding0.707
  Main evaluation result showing best variant outperforms many proprietary and open-source baselines of comparable or larger sizes.
• Beam search DTG achieves ending NLI Diversity of 5.35 vs nucleus sampling's higher performance, starting from -5.05finding0.707
  Confirms nucleus sampling produces more semantically diverse outputs than beam search