Pith. sign in

REVIEW 13 cited by

GSM-Infinite: How Do Your LLMs Behave over Infinitely Increasing Context Length and Reasoning Complexity?

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 2502.05252 v1 pith:5NF7I72B submitted 2025-02-07 cs.CL cs.AI

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

Long-context large language models (LLMs) have recently shown strong performance in information retrieval and long-document QA. However, to tackle the most challenging intellectual problems, LLMs must reason effectively in long and complex contexts (e.g., frontier mathematical research). Studying how LLMs handle increasing reasoning complexity and context length is essential, yet existing benchmarks lack a solid basis for quantitative evaluation. Inspired by the abstraction of GSM-8K problems as computational graphs, and the ability to introduce noise by adding unnecessary nodes and edges, we develop a grade school math problem generator capable of producing arithmetic problems with infinite difficulty and context length under fine-grained control. Using our newly synthesized GSM-Infinite benchmark, we comprehensively evaluate existing LLMs. We find a consistent sigmoid decline in reasoning performance as complexity increases, along with a systematic inference scaling trend: exponentially increasing inference computation yields only linear performance gains. These findings underscore the fundamental limitations of current long-context LLMs and the key challenges in scaling reasoning capabilities. Our GSM-Infinite benchmark provides a scalable and controllable testbed for systematically studying and advancing LLM reasoning in long and complex contexts.

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.

  1. StaminaBench: Stress-Testing Coding Agents over 100 Interaction Turns

    cs.SE 2026-06 unverdicted novelty 7.0 of 10

    StaminaBench evaluates coding agents over 100 procedurally generated change requests to a REST API, finding that tested models fail within 5-6 turns without feedback but improve up to 12x with test feedback and good h...

  2. LEAD: Breaking the No-Recovery Bottleneck in Long-Horizon Reasoning

    cs.AI 2026-03 unverdicted novelty 7.0 of 10

    LEAD lets LLMs solve checkers jumping puzzles up to size 13 by using lookahead to recover from irreversible errors on hard steps that break extreme decomposition.

  3. Copy Less, Ground More: Overcoming Repetitive Copying in Long-Context Reasoning via Evidence-Aware Reinforcement Learning

    cs.CL 2026-07 conditional novelty 6.0 of 10

    A reward that rewards n-gram overlap with key evidence and penalizes overlap with distractors improves long-context RL reasoning accuracy by up to +4.6 points while cutting thinking length.

  4. Bridging Structure and Language: Graph-Based Visual Reasoning for Autonomous Road Understanding

    cs.CV 2026-05 unverdicted novelty 6.0 of 10

    A graph-grounded Combined Road Substrate framework generates traceable QA pairs from road maps to improve small VLMs on compositional road reasoning tasks.

  5. The Power of Power Law: Asymmetry Enables Compositional Reasoning

    cs.AI 2026-04 unverdicted novelty 6.0 of 10

    Power-law data sampling creates beneficial asymmetry in the loss landscape that lets models acquire high-frequency skill compositions first, enabling more efficient learning of rare long-tail skills than uniform distr...

  6. The Power of Power Law: Asymmetry Enables Compositional Reasoning

    cs.AI 2026-04 unverdicted novelty 6.0 of 10

    Power-law data distributions outperform uniform ones for compositional reasoning by creating asymmetry that lets frequent skill compositions scaffold rare ones with less data.

  7. From Local Corrections to Generalized Skills: Improving Neuro-Symbolic Policies with MEMO

    cs.RO 2026-03 conditional novelty 6.0 of 10

    Clustering multi-user multi-task language corrections into a retrieval-augmented skillbook of code templates lets neuro-symbolic robots generate new skills and raise zero-shot success on novel tasks.

  8. Factored Causal Representation Learning for Robust Reward Modeling in RLHF

    cs.LG 2026-01 unverdicted novelty 6.0 of 10

    A factored causal representation learning method improves robustness of reward models in RLHF by isolating causal factors from biases like length and sycophancy using adversarial gradient reversal.

  9. vAttention: Verified Sparse Attention

    cs.LG 2025-10 conditional novelty 6.0 of 10

    vAttention is a sparse attention method that mixes heavy-hitter tokens with a statistically sized random sample to provide (ε, δ)-guaranteed approximation of full attention.

  10. LongReasonArena: A Long Reasoning Benchmark for Large Language Models

    cs.CL 2025-08 conditional novelty 6.0 of 10

    A benchmark where AI models must run long algorithmic computations reveals a linear drop in accuracy as the log of reasoning steps increases, with top models scoring 7-16% on the hardest level.

  11. Copy Less, Ground More: Overcoming Repetitive Copying in Long-Context Reasoning via Evidence-Aware Reinforcement Learning

    cs.CL 2026-07 conditional novelty 5.0 of 10

    GEAR adds a grounding reward for overlap with key evidence and a distractor penalty to the RL accuracy signal, reducing repetitive copying and improving long-context reasoning accuracy by up to 4.6 points over accurac...

  12. ATLAS: All-round Testing of Long-context Abilities across Scales

    cs.CL 2026-05 unverdicted novelty 5.0 of 10

    ATLAS is a length-dependent benchmarking framework that evaluates 26 models on 8 capability dimensions and shows substantial rank changes when moving from 128K to 1M token ranges.

  13. MiMo-V2-Flash Technical Report

    cs.CL 2026-01 unverdicted novelty 5.0 of 10

    MiMo-V2-Flash is a 309B/15B MoE model trained on 27T tokens with hybrid attention and multi-teacher on-policy distillation that matches larger models like DeepSeek-V3.2 while enabling 2.6x faster decoding via repurpos...

Pith tools