R637 โ€” SoftCtx ร— MidRank ร— LoBeta MegaExtra ep3 ร— LoLR (offline DPO)

Affine SN120 challenger trained to beat the live king on Reason v3 (teacher-anchored score: lpC(y_C|z_A) โˆ’ lpC(y_C|โˆ…)).

How this checkpoint was trained

  • Base / parent: unconst/Affine-5czsc2fc98-r252-merged@b42d6245d77fe30885ea8a90387771e1bc465e0f (our crowned r252, reign 33)
  • Method: offline DPO on Reason-ranked pairs (not SFT / not online GRPO)
  • What was optimized: preference for higher teacher-side Reason on mined pairs
  • Data: SoftCtx ร— MidRank ร— LoBeta pair set (soft context band, mid LoRA rank, low DPO ฮฒ). See plan.md / dpo_duel_reason.jsonl under mining/experiments/r637-r252-offline-dpo-hialpha-midrank-lobeta-softctx-megaextrasteps-ep3-lolr.
  • Key hyperparameters:
    • LoRA r=32 (MidRank), ฮฑ=128 (HiAlpha)
    • ฮฒ=0.02 (LoBeta)
    • lr=1e-6 (LoLR)
    • max_len=12288 (SoftCtx)
    • max_steps=3600 MegaExtra; epochs=3
  • Hardware: Lium mine-r226-marsplan-fullft-1 (brave) GPUs 2,3 train+merge; host-relay โ†’ n80 on mine-r262-kevin-v5-nonking-grpo-1 (golden) GPUs 6,7 โ†’ /tmp/r637_merged
  • Local n80 vs live king reign34 (cryptoDev23/Affine-5Dku3dYp9j-hk8161@55b7ffe0โ€ฆ):
    • margin +0.005735, SE 0.001973, z=2.91, n=77
    • bar max(2ยทSE, ฮด=0.002) = 0.003946 (~1.45ร—)
    • thought median 168.5 (โ‰ฅ80), B pass 0.4125 (โ‰ฅ0.30)
    • decision: Stage-5 licensed (r637_decision_reign34.json, p3643)
  • Lineage: amplify of R580 Soft MidRank LoBeta Mega ep1 ~0.39ร— after R610 ep2 @5e-6 REFUTE; โ‰  Soft MidRank MidBeta R635 ep3 LoLR REFUTE
  • Experiment path: mining/experiments/r637-r252-offline-dpo-hialpha-midrank-lobeta-softctx-megaextrasteps-ep3-lolr

Intended use

SN120 Affine miner submission / evalsrv Reason duel. Not a general chat model.

License

Follows base model + Affine mining artifacts policy.

Downloads last month
17
Safetensors
Model size
35B params
Tensor type
BF16
ยท
Inference Providers NEW
This model isn't deployed by any Inference Provider. ๐Ÿ™‹ Ask for provider support

Model tree for elevateecho/sn120-7aded176a1f1

Finetuned
(18)
this model