Pith. sign in

REVIEW 11 cited by

Faith and Fate: Limits of Transformers on Compositionality

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 2305.18654 v3 pith:BZKHMBUM submitted 2023-05-29 cs.CL cs.AIcs.LG

Faith and Fate: Limits of Transformers on Compositionality

classification cs.CL cs.AIcs.LG
keywords taskscompositionalreasoningllmsmodelsmulti-stepproblemstransformer
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved
0 comments
read the original abstract

Transformer large language models (LLMs) have sparked admiration for their exceptional performance on tasks that demand intricate multi-step reasoning. Yet, these models simultaneously show failures on surprisingly trivial problems. This begs the question: Are these errors incidental, or do they signal more substantial limitations? In an attempt to demystify transformer LLMs, we investigate the limits of these models across three representative compositional tasks -- multi-digit multiplication, logic grid puzzles, and a classic dynamic programming problem. These tasks require breaking problems down into sub-steps and synthesizing these steps into a precise answer. We formulate compositional tasks as computation graphs to systematically quantify the level of complexity, and break down reasoning steps into intermediate sub-procedures. Our empirical findings suggest that transformer LLMs solve compositional tasks by reducing multi-step compositional reasoning into linearized subgraph matching, without necessarily developing systematic problem-solving skills. To round off our empirical study, we provide theoretical arguments on abstract multi-step reasoning problems that highlight how autoregressive generations' performance can rapidly decay with\,increased\,task\,complexity.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 11 Pith papers

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

  1. A Verifiable Search Is Not a Learnable Chain-of-Thought

    cs.LG 2026-06 unverdicted novelty 7.0

    Verifiable search procedures cannot be learned as forward chain-of-thought by language models; they instead learn memorization, verification, or require precomputed catalogs.

  2. Proper Scoring Rules for Agentic Uncertainty Quantification

    cs.AI 2026-05 unverdicted novelty 7.0

    Introduces Trajectory Proper Score (TPS) as a strictly proper family of trajectory-level scoring rules that elicits the complete prefix-conditioned success probability process.

  3. Training Transformers as a Universal Computer

    cs.AI 2026-04 unverdicted novelty 7.0

    A transformer trained on random meaningless MicroPy programs generalizes to execute diverse human-written programs, providing empirical evidence it can act as a universal computer.

  4. TSVer: A Benchmark for Fact Verification Against Time-Series Evidence

    cs.CL 2025-11 unverdicted novelty 7.0

    TSVer is a new benchmark dataset for fact verification against time-series evidence, with 304 annotated real-world claims, 400 time series, verdicts, and justifications, plus baseline results showing current models struggle.

  5. Arithmetic Pedagogy for Language Models

    cs.CL 2026-06 unverdicted novelty 6.0

    A small GPT-2 model trained from scratch on GASING-derived CoT supervision for arithmetic reaches over 80% held-out accuracy, exhibits three learning phases, and develops both procedural and associative reasoning.

  6. When Should Users Check? Modeling Confirmation Frequency inMulti-Step Agentic AI Tasks

    cs.HC 2025-10 conditional novelty 6.0

    A decision-theoretic model based on the observed Confirmation-Diagnosis-Correction-Redo user pattern places intermediate confirmations in AI agent tasks, yielding 81% user preference and 13.54% faster completion versu...

  7. How Do Language Models Compose Functions?

    cs.CL 2025-10 conditional novelty 6.0

    LLMs solve compositional factual recall either by computing intermediates or directly, with mechanism choice correlated to translation geometry in embedding spaces.

  8. LiveCodeBench: Holistic and Contamination Free Evaluation of Large Language Models for Code

    cs.SE 2024-03 unverdicted novelty 6.0

    LiveCodeBench collects 400 recent contest problems to create a contamination-free benchmark evaluating LLMs on code generation and related capabilities like self-repair and execution.

  9. Game Theory Driven Multi-Agent Framework Mitigates Language Model Hallucination

    cs.AI 2026-07 conditional novelty 5.0

    A game-framed multi-agent system synthesizes large chemistry CoT/QA corpora and trains OmniChem-7B to near GPT-4o-mini performance with a large reported drop in hallucinations.

  10. Handling Feature Heterogeneity with Learnable Graph Patches

    cs.LG 2026-06 unverdicted novelty 5.0

    Learnable graph patches enable domain-agnostic pre-training of graph models by decomposing heterogeneous graphs into transferable semantic units via patch encoders and aggregators.

  11. Beyond Exponential Decay: Rethinking Error Accumulation in Large Language Models

    cs.CL 2025-05 unverdicted novelty 5.0

    LLM errors concentrate in sparse key tokens (5-10% of sequence) at semantic decision junctions, yielding a new reliability model that explains sustained long-context coherence.