REVIEW 3 cited by
What Can Neural Networks Reason About?
Not yet reviewed by Pith; the record is open.
This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.
SPECIMEN: schema-true, not a live event
T0 review · schema-true
One-sentence machine reading of the paper's core claim.
pith:XXXXXXXX · record.json · timestamp
read the original abstract
Neural networks have succeeded in many reasoning tasks. Empirically, these tasks require specialized network structures, e.g., Graph Neural Networks (GNNs) perform well on many such tasks, but less structured networks fail. Theoretically, there is limited understanding of why and when a network structure generalizes better than others, although they have equal expressive power. In this paper, we develop a framework to characterize which reasoning tasks a network can learn well, by studying how well its computation structure aligns with the algorithmic structure of the relevant reasoning process. We formally define this algorithmic alignment and derive a sample complexity bound that decreases with better alignment. This framework offers an explanation for the empirical success of popular reasoning models, and suggests their limitations. As an example, we unify seemingly different reasoning tasks, such as intuitive physics, visual question answering, and shortest paths, via the lens of a powerful algorithmic paradigm, dynamic programming (DP). We show that GNNs align with DP and thus are expected to solve these tasks. On several reasoning tasks, our theory is supported by empirical results.
Forward citations
Cited by 3 Pith papers
-
Distance-Preserving Embeddings in Inhomogeneous Random Graphs
On supercritical inhomogeneous random graphs, multi-scale landmark embeddings achieve (1±ε)-distortion of shortest paths at dimension Ω(n^{1-ε} log n), far below worst-case, with universal kernel extensions and transf...
-
G1: Teaching LLMs to Reason on Graphs with Reinforcement Learning
Reinforcement learning on synthetic graph-theoretic tasks sharply improves LLM graph reasoning, transferring to larger graphs, new encodings, and real-world tasks.
-
When More is Less: Understanding Chain-of-Thought Length in LLMs
LLM accuracy follows an inverted U in chain-of-thought length, with an optimal length that grows with task difficulty and shrinks with model capability.
Discussion (0). Continue with ORCID to comment.