ReasoningFlow: Semantic Structure of Complex Reasoning Traces
By: Jinu Lee , Sagnik Mukherjee , Dilek Hakkani-Tur and more
Potential Business Impact:
Helps computers think and learn better.
Large reasoning models (LRMs) generate complex reasoning traces with planning, reflection, verification, and backtracking. In this work, we introduce ReasoningFlow, a unified schema for analyzing the semantic structures of these complex traces. ReasoningFlow parses traces into directed acyclic graphs, enabling the characterization of distinct reasoning patterns as subgraph structures. This human-interpretable representation offers promising applications in understanding, evaluating, and enhancing the reasoning processes of LRMs.
Similar Papers
ReasonGraph: Visualisation of Reasoning Paths
Computation and Language
Shows how smart computer programs think step-by-step.
The Geometry of Reasoning: Flowing Logics in Representation Space
Artificial Intelligence
Shows how computers "think" through math.
Topology of Reasoning: Understanding Large Reasoning Models through Reasoning Graph Properties
Artificial Intelligence
Makes math AI better by understanding its thinking.