LatentChem: From Textual CoT to Latent Thinking in Chemical Reasoning
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , , , , , , , , , , , , , , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2026
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866915858420334592 |
|---|---|
| author | Ye, Xinwu Mao, Yicheng Zhang, Jia Liu, Yimeng Hao, Li Wu, Fang Li, Zhiwei Liao, Yuxuan Wang, Zehong Wu, Yingcheng Liu, Zhiyuan Yin, Zhenfei Yuan, Li Torr, Philip Sun, Huan Zeng, Xiangxiang Wang, Mengdi Cong, Le Gao, Shenghua Tang, Xiangru |
| author_facet | Ye, Xinwu Mao, Yicheng Zhang, Jia Liu, Yimeng Hao, Li Wu, Fang Li, Zhiwei Liao, Yuxuan Wang, Zehong Wu, Yingcheng Liu, Zhiyuan Yin, Zhenfei Yuan, Li Torr, Philip Sun, Huan Zeng, Xiangxiang Wang, Mengdi Cong, Le Gao, Shenghua Tang, Xiangru |
| contents | Chemical large language models (LLMs) predominantly rely on explicit Chain-of-Thought (CoT) in natural language to perform complex reasoning. However, chemical reasoning is inherently continuous and structural, and forcing it into discrete linguistic tokens introduces a fundamental representation mismatch that constrains both efficiency and performance. We introduce LatentChem, a latent reasoning interface that decouples chemical computation from textual generation, enabling models to perform multi-step reasoning directly in continuous latent space while emitting language only for final outputs. Remarkably, we observe a consistent emergent behavior: when optimized solely for task success, models spontaneously internalize reasoning, progressively abandoning verbose textual derivations in favor of implicit latent computation. This shift is not merely stylistic but computationally advantageous. Across diverse chemical reasoning benchmarks, LatentChem achieves a 59.88\% non-tie win rate over strong CoT-based baselines on ChemCoTBench, while delivering a 10.84$\times$ average reduction in reasoning overhead. Our results provide empirical evidence that chemical reasoning is more naturally and effectively realized as continuous latent dynamics rather than discretized linguistic trajectories. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2602_07075 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | LatentChem: From Textual CoT to Latent Thinking in Chemical Reasoning Ye, Xinwu Mao, Yicheng Zhang, Jia Liu, Yimeng Hao, Li Wu, Fang Li, Zhiwei Liao, Yuxuan Wang, Zehong Wu, Yingcheng Liu, Zhiyuan Yin, Zhenfei Yuan, Li Torr, Philip Sun, Huan Zeng, Xiangxiang Wang, Mengdi Cong, Le Gao, Shenghua Tang, Xiangru Chemical Physics Artificial Intelligence Computation and Language Machine Learning Chemical large language models (LLMs) predominantly rely on explicit Chain-of-Thought (CoT) in natural language to perform complex reasoning. However, chemical reasoning is inherently continuous and structural, and forcing it into discrete linguistic tokens introduces a fundamental representation mismatch that constrains both efficiency and performance. We introduce LatentChem, a latent reasoning interface that decouples chemical computation from textual generation, enabling models to perform multi-step reasoning directly in continuous latent space while emitting language only for final outputs. Remarkably, we observe a consistent emergent behavior: when optimized solely for task success, models spontaneously internalize reasoning, progressively abandoning verbose textual derivations in favor of implicit latent computation. This shift is not merely stylistic but computationally advantageous. Across diverse chemical reasoning benchmarks, LatentChem achieves a 59.88\% non-tie win rate over strong CoT-based baselines on ChemCoTBench, while delivering a 10.84$\times$ average reduction in reasoning overhead. Our results provide empirical evidence that chemical reasoning is more naturally and effectively realized as continuous latent dynamics rather than discretized linguistic trajectories. |
| title | LatentChem: From Textual CoT to Latent Thinking in Chemical Reasoning |
| topic | Chemical Physics Artificial Intelligence Computation and Language Machine Learning |
| url | https://arxiv.org/abs/2602.07075 |