The authors first propose Symbolic Closure Analysis (SCA), a theoretical lens characterizing how branching structures and sparse rewards induce exploration and compounding biases in long-horizon reasoning with local admissibility.
Topological guidance reduces long-horizon reasoning biases in LLMs
SAGE framework uses algebraic sparsification and hyperbolic embeddings to improve performance on sparse-reward reasoning tasks, achieving up to 8x gains on the Andrews-Curtis problem.
Top University
Xinyue Zeng · Jiawei Zhang · Yujun Yan · Dawei Zhou
Virginia Tech · University of Wisconsin Madison · Dartmouth College
Research Digest··2 min read
Zeng et al.
Why this paper
From Dartmouth College and 2 others · Released code · Part of World Model Planning for Agents, now 22 papers
In one line
SAGE mitigates long-horizon reasoning biases in LLMs via algebraic sparsification and hyperbolic structural guidance, achieving up to 8x improvement on Andrews-Curtis.
What it released
Code
What we could check
- ✓Code link in the paper (github.com)
- ·No weights link found
- ·No dataset link found
- ·No compute details found
- ·No stated limitations found
- ✓Reports numbers on named benchmarks
Observed from the paper text and links we have. Absence here means we did not find it, not that it does not exist.
§