Adapter Details

• Base model: Qwen/Qwen3.5-9B
• PEFT type: LORA
• Task type: CAUSAL_LM
• Adapter source path: stablecoin_phase1_pipeline/training/outputs/round3_conservative_repair/checkpoints/full_train_20260602T101206Z/adapter
• Formal eval report source: stablecoin_phase1_pipeline/evaluation/reports/qwen35_round3_formal_eval_report.md
• Formal eval result source: stablecoin_phase1_pipeline/evaluation/results/qwen35_round3_formal_eval_20260604T052411Z

Formal Evaluation

• Sample count: 300
• Pass rate: 0.9867
• Mean score: 0.8290
• Reasoning leak count: 0
• Missing citation count: 0
• Invalid citation ID count: 0
• Empty answer count: 0
• Material boundary error count: 0
• API key leak count: 0
• Format error count: 0
• Eval set SHA256: 1b18137c7c1a363c5f5bed2bbdd5f83d11cbe83a590671d211c134056a86861f

See formal_eval_report.md for the copied formal evaluation report.

Intended Use

This adapter is intended for controlled stablecoin-domain answer generation workflows where outputs are expected to remain grounded in provided materials and use strict citation behavior.

Limitations

This adapter is not legal advice. Outputs must be reviewed by qualified humans before use in legal, compliance, financial, or other high-stakes settings.

The adapter should be used with the same prompt, citation, and material-boundary controls used during evaluation. Performance outside that setup has not been established.