sxiong/MLR_concat10_output

MLR-Concat10 Output

This dataset contains MLR_model generated hierarchical reasoning traces for the MLR-Concat10 long-horizon reasoning benchmark. Each record contains one prompt with 10 independent questions, the structured reasoning trace produced while solving them in sequence, predicted answers, gold answers, and per-question correctness labels.

It accompanies the long-horizon reasoning setting in Enhancing Language Model Reasoning with Structured Multi-Level Modeling (ICLR 2026).

Configurations

Config	Source questions	Output records	Questions per record	Tested models	Accuracy
math500_qwen_1.5b	MATH500	10	10	MLR_executor_Qwen-1.5B, MLR_planner_Qwen-1.5B-LoRA	50%
aime24_qwen_1.5b	AIME24	10	10	MLR_executor_Qwen-1.5B, MLR_planner_Qwen-1.5B-LoRA	13%
mixture_qwen_1.5b	MATH500, AIME24, GPQA-diamond, and BoardGameQA-Hard	10	10	MLR_executor_Qwen-1.5B, MLR_planner_Qwen-1.5B-LoRA	41%

Data format

Each JSONL row has the following fields:

Field	Type	Description
ids	list[string]	Source-question identifiers, in prompt order.
problem	string	The concatenated prompt containing 10 numbered questions.
steps	list[object]	Structured reasoning trace across all questions. Each step contains step_id, cognitive_mode, subgoal, deliverable, success_criterion, and execution.
pred_answers	list[string]	Model predictions, one per question and aligned with ids.
gold_answers	list[string]	Reference answers, one per question and aligned with ids.
correct	list[bool]	Per-question correctness labels comparing pred_answers with gold_answers.

Loading

from datasets import load_dataset

ds = load_dataset("sxiong/MLR_concat10_output", "math500_qwen_1.5b", split="train")
example = ds[0]

trace = example["steps"]
predictions = example["pred_answers"]
references = example["gold_answers"]
correctness = example["correct"]

Citation

@inproceedings{xiong2026enhancing,
  title={Enhancing language model reasoning with structured multi-level modeling},
  author={Xiong, Siheng and Payani, Ali and Fekri, Faramarz},
  booktitle={International Conference on Learning Representations},
  volume={2026},
  pages={36557--36610},
  year={2026}
}