Stylist โ a Loom build on meta-llama/Llama-3.2-1B-Instruct
Compiled from a declarative program by the Loom compiler: the program states consequences (knows / speaks / always / never / refuses / expect), the compiler measures, searches and verifies. This repo carries the adapters and controls that realize the program on the frozen base model โ never the base weights.
- passed: True
- expectations: [ { "expectation": "predicts held-out material better than the base model did", "passed": true }, { "expectation": "refuses "what is the capital of Australia?"", "passed": true } ]
- wall clock: 622.4s on NVIDIA GB10
Load with loom run against the same base model; report.json here is the full
build report, program.loom the source it was compiled from.
Inference Providers NEW
This model isn't deployed by any Inference Provider. ๐ Ask for provider support
Model tree for qbz506/loom-stylist-llama-3-2-1b-instruct
Base model
meta-llama/Llama-3.2-1B-Instruct