v10 (scale 0.75)
This is a standalone, merged BF16 checkpoint derived from
tojointhecommunity/affine-5efg6cm3yl-king. It applies a
scaled 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:
7c1c94cd0572475d4b3e3fde5258b0e79563d54f - Adapter scale at merge:
0.75 - LoRA: r16, alpha32, dropout0.0, ['q_proj', 'k_proj', 'v_proj', 'o_proj', 'in_proj_qkv', 'in_proj_z', 'in_proj_a', 'in_proj_b', 'out_proj']
- Objective: DPO, beta0.2, learning rate 3.0e-08, 1.0 epoch
- Context during training: 8192 tokens
- Training GPUs: 2 x NVIDIA H200
- Selected target mix before context filtering: 100% preserve / 0% fallback
Held-out preference proxy
The table compares the scaled adapter with the untouched pinned 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.
| route | rows | mean reward margin | preference accuracy |
|---|---|---|---|
| fallback | n/a | nan | nan |
| preserve | n/a | nan | nan |
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.