Research map / Large language models / Agents, code and math
XDomainBench: Diagnosing Reasoning Collapse in High-Dimensional Scientific Knowledge Composition
Gong Zhiren, Tiantong Wu, Jiaming Zhang, Fuyao Zhang, CHE WANG, Yurong Hao, Yikun Hou, Foo Ping, Yilei Zhao, Fei Huang, Chau Yuen, Wei Yang Bryan Lim
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| increase composition order from 1 to 4 domains (ablation) | LLMs (large models mean) · XDomainBench deterministic | SessionSuccess@tau: 38.7 -> 27.1 | Hurt |
| increase composition order from 1 to 4 domains (ablation) | LLMs (large models mean) · XDomainBench deterministic | Recall: 28.8 -> 18.2 | Hurt |
| increase composition order from 1 to 4 domains (ablation) | Small models (mean) · XDomainBench deterministic | Recall: 26.2 -> 17.4 | Hurt |
| increase composition order from 1 to 4 domains (ablation) | MoE models (mean) · XDomainBench deterministic | Recall: 60.1 -> 35.2 | Hurt |
| increase composition order from 1 to 4 domains (ablation) | MoE models (mean) · XDomainBench deterministic | SessionSuccess@tau: 82.1 -> 42.0 | Hurt |
| score with semantic judge instead of token recall (ablation) | evaluation metric · XDomainBench, k=1 to 4 | accuracy: Judge 37.1 -> 26.8; Recall 31.3 -> 21.9 | No clear effect |
| use MoE models instead of dense large or small models | MoE models · XDomainBench deterministic, k=1 | SessionSuccess@tau: 82.1 vs 38.7 (large mean) | Improved |
| compose domains in turn 1 versus single-domain baselines (ablation) | LLMs · XDomainBench turn-1 | turn-1 correctness | Hurt |
| volatile difficulty/mixture patterns (fluctuate, spike) versus stable (ablation) | LLMs · XDomainBench sessions | session collapse rate | Hurt |
Closest papers by meaning
- SciBench: Evaluating College-Level Scientific Problem-Solving Abilities of Large Language Models (ICML 2024 · similarity 0.87)
- SciReason: A Controllable Benchmark for Scientific Reasoning in LLMs (NeurIPS 2026 · similarity 0.87)
- LongCoT: Benchmarking Long-Horizon Chain-of-Thought Reasoning (ICML 2026 · similarity 0.86)
- MDK12-Bench: A Multi-Discipline Benchmark for Evaluating Reasoning in Multimodal Large Language Models (AAAI 2026 · similarity 0.86)
- CURIE: Evaluating LLMs on Multitask Scientific Long-Context Understanding and Reasoning (ICLR 2025 · similarity 0.86)
- The Illusion of Thinking: Understanding the Strengths and Limitations of Reasoning Models via the Lens of Problem Complexity (NeurIPS 2025 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
