tinyLLM adapter: opd_ifeval_rank64
中文项目文档 · English project overview · Live demo
Base model: chris0809/tinyLLM-0.51B-SFT
This native tinyLLM LoRA was trained with on-policy distillation from MiniCPM3-4B on AllenAI RLVR-IFEval. Google IFEval strict instruction accuracy improved from 23.74% to 25.18% (+1.44 percentage points).
from transformers import AutoModelForCausalLM
model = AutoModelForCausalLM.from_pretrained("chris0809/tinyLLM-0.51B-SFT", trust_remote_code=True)
model.load_lora_pretrained("chris0809/tinyLLM-0.51B-IFEval-OPD")
Google IFEval is evaluated with its original instructions, so the reported
score does not prepend a reasoning prompt. For a separate mathematics or
multi-step task, the shared base tokenizer also supports
tokenizer.apply_chat_template(..., enable_thinking=True).
The adapter uses tinyLLM's native multi-adapter implementation and is stored with safetensors; it is not a PEFT adapter.
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support
Model tree for chris0809/tinyLLM-0.51B-IFEval-OPD
Base model
chris0809/tinyLLM-0.51B-SFT