Pith. sign in

REVIEW 13 cited by

ZebraLogic: On the Scaling Limits of LLMs for Logical Reasoning

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 2502.01100 v2 pith:DFVYWZHK submitted 2025-02-03 cs.AI cs.CLcs.LG

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

We investigate the logical reasoning capabilities of large language models (LLMs) and their scalability in complex non-monotonic reasoning. To this end, we introduce ZebraLogic, a comprehensive evaluation framework for assessing LLM reasoning performance on logic grid puzzles derived from constraint satisfaction problems (CSPs). ZebraLogic enables the generation of puzzles with controllable and quantifiable complexity, facilitating a systematic study of the scaling limits of models such as Llama, o1 models, and DeepSeek-R1. By encompassing a broad range of search space complexities and diverse logical constraints, ZebraLogic provides a structured environment to evaluate reasoning under increasing difficulty. Our results reveal a significant decline in accuracy as problem complexity grows -- a phenomenon we term the curse of complexity. This limitation persists even with larger models and increased inference-time computation, suggesting inherent constraints in current LLM reasoning capabilities. Additionally, we explore strategies to enhance logical reasoning, including Best-of-N sampling, backtracking mechanisms, and self-verification prompts. Our findings offer critical insights into the scalability of LLM reasoning, highlight fundamental limitations, and outline potential directions for improvement.

Discussion (0). Sign in to comment.

Forward citations

Cited by 13 Pith papers

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

  1. Toward Skill-Native LLMs: Skill Entropy for Benchmarking and Training Long-Horizon Reasoning

    cs.CL 2026-08 conditional novelty 7.0 of 10

    Skill entropy, a reference-model-based measure of skill-switching difficulty, calibrates a new cross-skill benchmark and serves as an RL reward, more than doubling small models' scores.

  2. Enigmata: Scaling Logical Reasoning in Large Language Models with Synthetic Verifiable Puzzles

    cs.CL 2025-05 conditional novelty 7.0 of 10

    Enigmata's synthetic puzzles with verifiable rewards lift a 32B model to 32.8% on ARC-AGI, above o3-mini-high and o1, and give small apparent gains on math and STEM when added to Seed1.5-Thinking.

  3. Reasoning Core: Designing Broad Procedural Data for Completion-Supervised Reasoning Training

    cs.CL 2026-08 conditional novelty 6.0 of 10

    A new collection of 50 procedural generators designed for completion-supervised fine-tuning beats three existing procedural collections and a no-procedural baseline on reasoning benchmarks at 3B scale in mean scores.

  4. The Format Tax

    cs.CL 2026-04 conditional novelty 6.0 of 10

    Structured-output instructions alone impose a large accuracy tax on open-weight LLMs; decoupling freeform reasoning from formatting recovers most of it, while recent closed models largely avoid the tax.

  5. Estimating the Empowerment of Language Model Agents

    cs.AI 2025-09 conditional novelty 6.0 of 10

    EELMA estimates the mutual information between an LM agent's actions and future text states, and this 'empowerment' is shown to correlate with task performance across toy games and WebArena.

  6. Rethinking Reasoning Quality in Large Language Models through Enhanced Chain-of-Thought via RL

    cs.AI 2025-09 conditional novelty 6.0 of 10

    DRER rewards CoT trajectories that increase the model's likelihood of the correct answer, plus a length penalty, and the new LogicTree benchmark reportedly lifts a 7B model's average accuracy from 0.13 to 0.60.

  7. Can One Domain Help Others? A Data-Centric Study on Multi-Domain Reasoning via Reinforcement Learning

    cs.AI 2025-07 conditional novelty 6.0 of 10

    Training a Qwen2.5-7B model with GRPO on math and puzzle data improves both domains, code transfer depends on the starting model, and template or reward mismatches sharply hurt performance.

  8. Reasoning Strategies in Large Language Models: Can They Follow, Prefer, and Optimize?

    cs.CL 2025-07 conditional novelty 6.0 of 10

    Prompting LLMs with distinct reasoning strategies and ensembling their outputs improves accuracy on logical deduction tasks, though not as consistently as the paper claims.

  9. Teaching LLM to Reason: Reinforcement Learning from Algorithmic Problems without Code

    cs.CL 2025-07 conditional novelty 6.0 of 10

    TeaR uses GRPO reinforcement learning on test-case output prediction for algorithmic problems, with no code shown, and reports broad reasoning gains across 17 benchmarks.

  10. LogiPlan: A Structured Benchmark for Logical Planning and Relational Reasoning in LLMs

    cs.AI 2025-06 conditional novelty 6.0 of 10

    LogiPlan introduces a three-task benchmark with controllable graph complexity, showing that although reasoning models excel at plan generation, all models degrade sharply on cycle detection and deep comparison questions.

  11. LogicPuzzleRL: Cultivating Robust Mathematical Reasoning in LLMs via Reinforcement Learning

    cs.LG 2025-06 reject novelty 6.0 of 10

    Reinforcement learning on seven verifiable logic puzzles improves LLM accuracy on out-of-distribution math benchmarks, especially mid-difficulty problems.

  12. Towards Fundamental Language Models: Does Linguistic Competence Scale with Model Size?

    cs.CL 2025-09 reject novelty 5.0 of 10

    Across 23 models from 135M to 32B parameters, internal factual knowledge scales about twice as fast with model size as linguistic competence, supporting modular small-model-plus-retrieval systems.

  13. AgentScope 1.0: A Developer-Centric Framework for Building Agentic Applications

    cs.AI 2025-08 unverdicted novelty 4.0 of 10

    AgentScope 1.0 packages the components needed to build, evaluate, and deploy LLM agent applications into one developer framework.

Pith tools