Wenxiang Geng
1 indexed paper
Recent (6 mo)
1With code
0Influential cites
0Benchmarked
0Publications per year
126
Top categories
ML×1AI×1
Frequent co-authors
Research Timeline
2026
The Paradox of Outcome Optimization: A Causal Information-Theoretic Bound on Reasoning Shortcuts in LLMs
The paper theoretically explains that optimizing LLMs solely on outcomes leads to brittle reasoning (Reward-Induced Manifold Collapse) by favoring low-complexity shortcuts, and proposes process-based supervision to fix this.
Highlighted terms show continued research focus across papers
Papers
cs.LGcs.AIRecentMay 30, 2026
The Paradox of Outcome Optimization: A Causal Information-Theoretic Bound on Reasoning Shortcuts in LLMs
Zihan Chen, Yiming Zhang, Wenxiang Geng, Zenghui Ding +1 more
The paper theoretically explains that optimizing LLMs solely on outcomes leads to brittle reasoning (Reward-Induced Manifold Collapse) by favoring low-complexity shortcuts, and proposes process-based…
View →