gold24k/v1

Affine R1032 selective fallback SF1 (scale 0.50)

This is a standalone, merged BF16 checkpoint derived from `unconst/Affine-5czsc2fc98-r1032-vera-odpo-midrank-hibeta-shortctx-ultraextra-ep4-midlr-merged`. It applies a half-strength selective-fallback LoRA trained on preserved positive turns and sanitized, task-specific alternatives for high-confidence negative turns. It does not require a runtime router, custom Python code, or a PEFT adapter.

Training summary

• —Exact parent revision: 62dfb322fdce5873543bd92692ab4ecc3e13f941
• —Adapter scale at merge: 0.50
• —LoRA: r16, alpha64, dropout0, all linear layers
• —Objective: DPO, beta0.2, learning rate 2e-7, one epoch
• —Context during training: 8192 tokens
• —Training GPUs: 2 x NVIDIA H200
• —Selected target mix before context filtering: 80% preserve / 20% fallback

Held-out preference proxy

The table compares the scaled adapter with the untouched R1032 parent. These small held-out metrics selected the merge strength; they are not a substitute for the exact full Affine duel on the dedicated evaluator.

Qualification status

Experimental candidate. Before submission, run exact stock-vLLM Affine duels, the exploit-pattern audit, repository preflight, and the official submission client check. selective_fallback_provenance.json contains the machine-readable training and merge record.