Evaluating the Architectural Reasoning Capabilities of LLM Provers via the Obfuscated Natural Number Game

• URL: http://arxiv.org/abs/2605.00677v1
• Date: Fri, 01 May 2026 14:03:05 GMT
• Title: Evaluating the Architectural Reasoning Capabilities of LLM Provers via the Obfuscated Natural Number Game
• Abstract summary: This paper identifies Architectural Reasoning as the necessary ability for future automated theorem discovery AI.<n>By renaming identifiers in the Natural Number Game in Lean 4, we created a zero-knowledge, closed environment.<n>We evaluate state-of-the-art models, finding a universal latency tax where obfuscation increases inference time.
• Score: 0.0
• License: http://creativecommons.org/publicdomain/zero/1.0/
• Abstract: While Large Language Models have achieved notable success on formal mathematics benchmarks such as MiniF2F, it remains unclear whether these results stem from genuine logical reasoning or semantic pattern matching against pre-training data. This paper identifies Architectural Reasoning: the ability to synthesize formal proofs using exclusively local axioms and definitions within an alien math domain, as the necessary ability for future automated theorem discovery AI. We use the Obfuscated Natural Number Game, a benchmark to evaluate Architectural Reasoning. By renaming identifiers in the Natural Number Game in Lean 4, we created a zero-knowledge, closed environment. We evaluate state-of-the-art models, finding a universal latency tax where obfuscation increases inference time. The results also reveal a divergence in robustness: while general models (Claude-Sonnet-4.5, GPT-4o) suffer performance degradation, reasoning models (DeepSeek-R1, GPT-5, DeepSeek-Prover-V2) maintain the same accuracy despite the absence of semantic cues. These findings provide a quantitative metric for assessing the true capacity for mathematical reasoning.

Related papers

• Benchmarking Testing in Automated Theorem Proving [39.65133452374143]
  T is a framework that evaluates the semantic correctness of formal theorems.<n>We construct a benchmark from 5 real-world Lean 4 repositories.<n>Experiments demonstrate that while state-of-the-art models achieve high compilation success, they perform significantly worse under our semantic metric.
  arXiv  Detail & Related papers  (2026-04-26T13:24:20Z)
• LiveMathematicianBench: A Live Benchmark for Mathematician-Level Reasoning with Proof Sketches [61.30693283718321]
  We present LiveMathematicianBench, a dynamic multiple-choice benchmark for research-level mathematical reasoning.<n>By grounding evaluation in newly published theorems, it provides a realistic testbed beyond memorized patterns.<n>It employs a proof-sketchguided distractor pipeline that uses high-level proof strategies to construct plausible but invalid answer choices.
  arXiv  Detail & Related papers  (2026-04-02T08:22:17Z)
• Pramana: Fine-Tuning Large Language Models for Epistemic Reasoning through Navya-Nyaya [0.0]
  Large language models produce fluent text but struggle with systematic reasoning.<n>When Apple researchers added irrelevant context, LLM performance degraded by 65%.<n>This inability to ground claims in traceable evidence limits AI reliability in domains requiring justification.<n>We introduce fine-tuning on Pramana, a 2,500-year-old Indian reasoning framework.
  arXiv  Detail & Related papers  (2026-02-14T23:45:29Z)
• Are Language Models Efficient Reasoners? A Perspective from Logic Programming [109.47572890883248]
  Modern language models (LMs) exhibit strong deductive reasoning capabilities, yet standard evaluations emphasize correctness while overlooking a key aspect of human-like reasoning: efficiency.<n>We propose a framework for assessing LM reasoning efficiency through the lens of logic programming.
  arXiv  Detail & Related papers  (2025-10-29T15:30:31Z)
• Do LLMs Overthink Basic Math Reasoning? Benchmarking the Accuracy-Efficiency Tradeoff in Language Models [6.312798900093575]
  Large language models (LLMs) achieve impressive performance on complex mathematical benchmarks yet sometimes fail on basic math reasoning.<n>This paper focuses on the fundamental tradeoff between accuracy and overthinking.<n>We introduce the Overthinking Score, a harmonic-mean metric combining accuracy and token-efficiency for holistic model evaluation.
  arXiv  Detail & Related papers  (2025-07-05T12:31:17Z)
• Safe: Enhancing Mathematical Reasoning in Large Language Models via Retrospective Step-aware Formal Verification [56.218970738892764]
  Chain-of-Thought prompting has become the de facto method to elicit reasoning capabilities from large language models (LLMs)<n>To mitigate hallucinations in CoT that are notoriously difficult to detect, current methods operate as opaque boxes and do not provide checkable evidence for their judgments, possibly limiting their effectiveness.<n>We propose a retrospective, step-aware formal verification framework $Safe$. Rather than assigning arbitrary scores, we strive to articulate mathematical claims in formal mathematical language Lean 4 at each reasoning step and provide formal proofs to identify hallucinations.
  arXiv  Detail & Related papers  (2025-06-05T03:16:08Z)
• FormalMATH: Benchmarking Formal Mathematical Reasoning of Large Language Models [17.919212265668783]
  We present FormalMATH, a large-scale Lean4 benchmark comprising 5,560 formally verified problems spanning from high-school Olympiad challenges to undergraduate-level theorems.<n>We introduce a novel human-in-the-loop autoformalization pipeline that integrates statement autoformalization, semantic verification, and negation-based disproof filtering strategies.<n>Our evaluation of state-of-the-art LLM-based theorem provers reveals significant limitations.
  arXiv  Detail & Related papers  (2025-05-05T15:37:00Z)
• Alchemy: Amplifying Theorem-Proving Capability through Symbolic Mutation [71.32761934724867]
  This work proposes Alchemy, a framework for data synthesis that constructs formal theorems through symbolic mutation.<n>For each candidate theorem in Mathlib, we identify all invocable theorems that can be used to rewrite or apply to it.<n>As a result, our method increases the number of theorems in Mathlib by an order of magnitude, from 110k to 6M.
  arXiv  Detail & Related papers  (2024-10-21T08:04:21Z)
• Lean-STaR: Learning to Interleave Thinking and Proving [53.923617816215774]
  We present Lean-STaR, a framework for training language models to produce informal thoughts prior to each step of a proof.<n>Lean-STaR achieves state-of-the-art results on the miniF2F-test benchmark within the Lean theorem proving environment.
  arXiv  Detail & Related papers  (2024-07-14T01:43:07Z)
• LINC: A Neurosymbolic Approach for Logical Reasoning by Combining Language Models with First-Order Logic Provers [60.009969929857704]
  Logical reasoning is an important task for artificial intelligence with potential impacts on science, mathematics, and society. In this work, we reformulating such tasks as modular neurosymbolic programming, which we call LINC. We observe significant performance gains on FOLIO and a balanced subset of ProofWriter for three different models in nearly all experimental conditions we evaluate.
  arXiv  Detail & Related papers  (2023-10-23T17:58:40Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.