Text Ranking
PEFT
Safetensors
English
reranker
cross-encoder
instruction-following
knowledge-graph
steerable
lora
qwen3
Eval Results (legacy)
Instructions to use Hanno-Labs/bosun-4b with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- PEFT
How to use Hanno-Labs/bosun-4b with PEFT:
from peft import PeftModel from transformers import AutoModelForCausalLM base_model = AutoModelForCausalLM.from_pretrained("Qwen/Qwen3-Reranker-4B") model = PeftModel.from_pretrained(base_model, "Hanno-Labs/bosun-4b") - Notebooks
- Google Colab
- Kaggle
| { | |
| "base_model": "Qwen/Qwen3-Reranker-4B", | |
| "system": "Judge whether the Document meets the requirements based on the Query and the Instruct provided. Note that the answer can only be \"yes\" or \"no\".", | |
| "prefix": "<|im_start|>system\nJudge whether the Document meets the requirements based on the Query and the Instruct provided. Note that the answer can only be \"yes\" or \"no\".<|im_end|>\n<|im_start|>user\n", | |
| "suffix": "<|im_end|>\n<|im_start|>assistant\n<think>\n\n</think>\n\n", | |
| "yes_id": 9693, | |
| "no_id": 2152, | |
| "max_len": 3072, | |
| "score": "sigmoid(logit_yes - logit_no)", | |
| "target": "soft", | |
| "per_row_instruct": true, | |
| "warrant_instruct": "Rate how strongly the two findings in the Document support the structural claim in the Query as a research lead. Lean \"yes\" if the claim is plausible and not contradicted \u2014 it could be true and more evidence would likely confirm it; lean \"no\" if a finding contradicts it, mischaracterizes the evidence, or the claim is a non-sequitur with no plausible path from these findings." | |
| } |