QLoRA fine-tune of Qwen3.5-4B for the MetroLLM-Bench transit-kiosk task. v24 is the leakage-free retraining used in the MetroLLM-Bench paper (teacher traces drawn only from the 717-case training partition; 238 cases held out). Supersedes continker/Qwen3.5-4B-metro-v23.

Held-out results (n=238, mean of 2 seeds)

Metric	4B base	4B + v24	Δ
Tier-1	89.32	91.32	+2.00
Composite	87.25	89.12	+1.87

The efficiency headline of the paper: at 2.6 GB Q4_K_M, the 4B student ties GPT-5.4 full at maximum reasoning effort on held-out tier-1 (91.32 vs 91.37).

adapter/ — LoRA adapter (rank 16, α 32; QLoRA 4-bit NF4) + tokenizer + chat template
Qwen3.5-4B-metro-v24-Q4_K_M.gguf — merged GGUF (2.6 GB)
training_summary.json

GGUF: `llama-server --hf-repo continker/Qwen3.5-4B-metro-v24 --hf-file Qwen3.5-4B-metro-v24-Q4_K_M.gguf`. The LoRA adapter keys use the `.language_model.` prefix; strip it to load onto text-only Qwen3.5-4B. Apache 2.0.

Citation

bibtex

@techreport{hendriks2026metrollm,
  title       = {MetroLLM-Bench: Evaluating Language Models as Transit Kiosk Runtimes},
  author      = {Hendriks, Remco},
  institution = {Continker},
  type        = {Technical report},
  number      = {v1},
  year        = {2026},
  month       = {8},
  doi         = {10.5281/zenodo.21893944},
  url         = {https://github.com/continker/metrollm-bench/blob/main/paper.pdf}
}

continker/
Qwen3.5-4B-metro-v24

Qwen3.5-4B-metro-v24

Held-out results (n=238, mean of 2 seeds)

Contents

Links

Citation