Pith. sign in

REVIEW 3 cited by

On the Paradox of Learning to Reason from Data

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

arxiv 2205.11502 v2 pith:3PIU7SKT submitted 2022-05-23 cs.CL cs.AI

classification cs.CLcs.AI
keywords learningreasoninglogicalbertdatafeaturesotherreason
verification ladder T0 review T1 audit T2 compute T3 formal
0 comments
read the original abstract

Logical reasoning is needed in a wide range of NLP tasks. Can a BERT model be trained end-to-end to solve logical reasoning problems presented in natural language? We attempt to answer this question in a confined problem space where there exists a set of parameters that perfectly simulates logical reasoning. We make observations that seem to contradict each other: BERT attains near-perfect accuracy on in-distribution test examples while failing to generalize to other data distributions over the exact same problem space. Our study provides an explanation for this paradox: instead of learning to emulate the correct reasoning function, BERT has in fact learned statistical features that inherently exist in logical reasoning problems. We also show that it is infeasible to jointly remove statistical features from data, illustrating the difficulty of learning to reason in general. Our result naturally extends to other neural models and unveils the fundamental difference between learning to reason and learning to achieve high performance on NLP benchmarks using statistical features.

Discussion (0). Sign in to comment.

Forward citations

Cited by 3 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Propositional Logic for Probing Generalization in Neural Networks

    cs.LG 2025-06 conditional novelty 6.0 of 10

    Standard neural architectures generalize to unseen variable and operator combinations, but systematically fail when negation is applied to an operator that was hidden during training.

  2. Mitigating Spurious Correlations in LLMs via Causality-Aware Post-Training

    cs.LG 2025-06 conditional novelty 5.0 of 10

    Fine-tuning a 3B LLM on randomly symbolized reasoning questions reduces spurious-correlation failures and improves OOD accuracy on CLadder and PrOntoQA.

  3. Grounding Natural Language for Multi-agent Decision-Making with Multi-agentic LLMs

    cs.AI 2025-08 unverdicted novelty 3.0 of 10

    A design framework for multi-agent LLMs combining prompting, memory, multimodal input, and fine-tuning, with promised ablations on social-dilemma games that are absent from the supplied text.

Pith tools