MetaDecision-31B-Rain

Decision service for JevBench. Base model: google/gemma-4-31B-it, used through Google's official quantization-aware-training (QAT) releases of it. The model has two parts that are loaded together and not merged; both are the files of MeTask-Rain/metask-jev-rain-31B-trained, unchanged and pinned by SHA-256 in MODEL_MANIFEST.json of the service code. This entry uses the same weights as that entry and changes only the serving rules.

Part	Location in that model repository	What it is
Base	repository root	The files of google/gemma-4-31B-it-qat-w4a16-ct at revision 52f3f65bc7a02d555763bc923bd1d9094898219d, byte for byte: Google's official compressed-tensors W4A16 checkpoint (4-bit weights, group size 32, symmetric).
Adapter	adapter/	Our LoRA (PEFT format): rank 16, alpha 32, on the language model's q/k/v/o/gate/up/down projections, trained by us in BF16 on google/gemma-4-31B-it-qat-q4_0-unquantized at revision 1e4d8beecacb8b7590c1d8bedd7335f687bf311f.

adapter/adapter_model.safetensors contains the trained checkpoint tensors unchanged plus, for the 10 full-attention layers (5, 11, …, 59) that compute values with the key projection, a v_proj LoRA that is a byte copy of the trained k_proj LoRA (20 tensors), so that vLLM applies the update to keys and values as in training.

Serving

The service code (service/ in the Hugging Face repository MeTask-Rain/MetaDecision-31B-Rain, mirrored at https://github.com/chenyuwang166-oss/MetaDecision-31B-Rain) starts vLLM 0.31.0 on a folder holding the files of that weight repository, with enable_lora, max_lora_rank=16, and the adapter/ folder as the single LoRA, applied to every request. Each decision is read from the next-token log-probabilities of the answer forms in one prompt pass; up to 49% of decisions then take a second, short drafting pass (at most 192 tokens) under a 9 s deadline, a choice decision answered in one pass although its length allowed the drafting pass is read a second time with the options in reverse order when its first-pass margin is below 4, and the reported probabilities get a per-type post-calibration (DISCLOSURE.md).

Equivalent vLLM usage of the model:

from vllm import LLM
from vllm.lora.request import LoRARequest
llm = LLM(model=MODEL_DIR, dtype="bfloat16", enable_lora=True, max_lora_rank=16)
adapter = LoRARequest("adapter", 1, MODEL_DIR + "/adapter")
# llm.generate(prompts, sampling_params, lora_request=adapter)

Facts on data and the serving rules are in DISCLOSURE.md of the service code (service/DISCLOSURE.md in the Hugging Face repository).

Contact: chenyuwang166@gmail.com