You need to agree to share your contact information to access this model

This repository is publicly accessible, but you have to accept the conditions to access its files and content.

YAML Metadata Warning:empty or missing yaml metadata in repo card

Check out the documentation for more information.

k128-roundc2-mix_r8x0.33-r10x0.33-r12x0.33-albedo-56-margin_p0.034-p0.015-merged

MERGED model (the submitted bytes, in the king's weight layout) of campaign round c2 against king v128 (base: dendriteholdings/albedo-qwen3.6-35b-king-CXXVIII with its row/column scaling undone - ft/king_normalize.py normalize; merge there, then king_normalize.py apply to return to the king's weight layout). Merge onto the base with peft to rebuild the submitted export; scale lora_alpha in adapter_config.json to scale the delta.

hotkey: albedo-56 (uid 207)
recipe: mix_r8x0.33-r10x0.33-r12x0.33
live result: LIVE PRE-EVAL PASSED; DUEL 1 WON +0.0341 (0.7614 vs 0.7272) at 01:09, CONFIRMATION +0.0145 (0.7700 vs 0.7555) at 01:38 UTC 6 Oct: not crowned (needs +0.025 twice); mean +0.024 under GLM-5.3-flash; albedo-56 spent
model: king128/export_roundc2: COMBINED ADAPTER (combine_adapters.py, exact rank-stacked sum on the clean v128 base): 1/3 each of rounds 8, 10, 12
gates: test submissions under the new judge (the user: make several submissions for test using our adapters combine)
eval run: 59c3f0c6

Downloads last month: 7

Safetensors

Model size

36B params

Tensor type

BF16

Inference Providers NEW

This model isn't deployed by any Inference Provider. 🙋 Ask for provider support