{"enabled":true,"releaseGates":true,"models":["qwen/qwen3.8-27b","qwen/qwen-2.5-7b-instruct"],"routerModels":["qwen/qwen3.8-27b","qwen/qwen-2.5-7b-instruct","phala/gemma-4-26b-a4b-uncensored"],"impactTerms":"Data Impact Trials v1. One aggregate scorecard comparing the same model with and without an approved private tool. The tool receives only the typed case inputs, and the assisted model receives only its decision. 70% of USDC pays the contributor and 30% pays Data Loom. Case inputs, reference answers, tool source files, and individual outputs are excluded from the scorecard.","terms":"Blind Model Trials v1. One aggregate benchmark scorecard, scored by Unicode-normalized, case-insensitive exact answer match with collapsed whitespace. No case prompts, reference answers, individual model outputs, download or training rights. Each model receives prompts, never reference answers. Verified Phala inference receipts are required for every response. 70% of USDC pays the contributor; 30% pays Data Loom. Results are a single run, not a guarantee of model quality. Already prepared purchases remain payable after an offer is disabled.","poolTerms":"Private Benchmark Pools v1. One aggregate scorecard from multiple opted-in dataset versions, with normalized exact-match scoring. No individual outputs, case prompts, answer keys, file access or training rights. Each model receives prompts, never reference answers. Every response requires a verified Phala receipt. The total price is the sum of the selected benchmark prices, up to 1,000 USDC. Contributors split 70% by case count; 30% goes to Data Loom. Already prepared payments remain valid after opt-out.","offers":[],"blockers":[]}