Pinned Loading

1. perfectman perfectman Public

   Social simulation experiment with the goal is to make AI personas feel like people: noticing unevenly, replying late, lurking, masking, forming private alliances, misreading silence, and creating e…

   TypeScript 10 3

2. habeas habeas Public

   Fine-tune Qwen3.8-27B to validate Form I-9 compliance (M-274/8 CFR 274a.2), citing violations. Synthetic-only data.

   Python 6

3. lossbench lossbench Public

   Expected-loss evaluation and control for agents that touch money. Severity-weighted, trajectory-level benchmark (reconciliation, payment repair, settlement risk) + record->calibrate->decide->escala…

   Python 4

4. reconforge reconforge Public

   Benchmark-grade evaluation for financial back-office reconciliation agents: a locally fine-tuned Qwen3-1.7B (MLX-LoRA) that beats DeepSeek on severity-weighted recall, measured on a reproducible AR…

   Python 4

5. plumb plumb Public

   AIA G702/G703 pay-app review. Path A curriculum audit: oracle-gated, contamination-controlled, 1.7B text policy.

   Python 5

6. suture suture Public

   Fine-tune a VL Qwen on insurance policy-issuance QC: binder vs issued policy images in, structured FLAG/PASS out. Seeded synthetic + verifier-as-oracle + SFT/ReST-EM.

   Python 5