### Request: add `IntelligenceLab/Long-Horizon-Terminal-Bench` to the benchmark allow-list

[zli12321]

Hi Hugging Face team,

Following the Evaluation Results docs, I've set up our dataset as a benchmark and would like to request that it be added to the current benchmark allow-list.

Dataset: https://huggingface.co/datasets/IntelligenceLab/Long-Horizon-Terminal-Bench

Benchmark configuration

eval.yaml is in the dataset repo root and validates successfully on the Hub.

• evaluation_framework: harbor
  • same framework already used by harborframework/terminal-bench-2.0
  • we are not requesting a new framework registration
• benchmark task:
  • id: lhtb
  • config: tasks
  • split: test

What this benchmark is

LHTB (Long-Horizon-Terminal-Bench) is a 46-task benchmark for evaluating how well LLM agents can sustain useful work in a containerized terminal over hundreds of steps.

Each task is packaged as a Harbor / Terminal-Bench 2.0 task folder, and grading is based on hidden rebuild-from-artifact verifiers rather than self-reported progress.

Relevant links:

• Dataset: https://huggingface.co/datasets/IntelligenceLab/Long-Horizon-Terminal-Bench
• Paper: https://arxiv.org/abs/2607.08964
• Code: https://github.com/zli12321/LHTB
• Project / public leaderboard: https://zli12321.github.io/LHTB/

Existing eval results

We already have community .eval_results PRs opened against model repos that reference this benchmark with task_id: lhtb, so leaderboard entries are ready once the dataset is allow-listed:

• MiniMaxAI/MiniMax-M3 discussion Improve right-to-left language support in inputs and outputs #24
• deepseek-ai/DeepSeek-V4-Pro discussion fixes a typo ffmeg --> ffmpeg #210
• moonshotai/Kimi-K2.6 discussion Different background color for markdown inline <code> blocks #55
• moonshotai/Kimi-K2.7-Code discussion Widget for Table to text generation #13
• tencent/Hy3 discussion Widget for zero-shot image classification #19
• zai-org/GLM-5.1 discussion Tracking integration for Image to Image #42
• zai-org/GLM-5.2 discussion Tracking integration for Table to Text generation #47

Request

Could you please add IntelligenceLab/Long-Horizon-Terminal-Bench to the benchmark allow-list so that the Benchmark badge and leaderboard can render on the dataset page?

Happy to provide any additional information if helpful. Thanks very much!

cc @julien-c @NathanHB

[NathanHB]

Hey! Sorry for the delay on this, just added you to the list. Super excited to see this benchmark on the hub, Would be great to see new sota models as they come out!!