Artificial Intelligence · arXiv:2603.19514 v2 · submitted 19 Mar 2026
Learning to Disprove: Formal Counterexample Generation with Large Language Models
The paper says
“Experiments on three newly collected benchmarks validate the advantages of our approach, showing that the mutation strategy and training framework yield substantial performance gains.”
from the abstract · picked out as the main claim by a classifier
Reviews
Reviews by people
No person has reviewed this paper on Pith yet. A scientist listed on Pith can review it for a fee, and their review is published here whatever it concludes.
The paper
Abstract
Mathematical reasoning demands two critical, complementary skills: constructing rigorous proofs for true statements and discovering counterexamples that disprove false ones. However, current AI efforts in mathematics focus almost exclusively on proof construction, often neglecting the equally important task of finding counterexamples. In this paper, we address this gap by fine-tuning large language models (LLMs) to reason about and generate counterexamples. We formalize this task as formal counterexample generation, which requires LLMs not only to propose candidate counterexamples but also to produce formal proofs that can be automatically verified in the Lean 4 theorem prover. To enable effective learning, we introduce a symbolic mutation strategy that constructs candidate counterexample-generation tasks by systematically extracting theorems and removing selected hypotheses. Together with curated datasets, this strategy enables a multi-reward expert iteration framework that substantially enhances both the effectiveness and efficiency of training LLMs for counterexample generation and theorem proving. Experiments on three newly collected benchmarks validate the advantages of our approach, showing that the mutation strategy and training framework yield substantial performance gains.
The paper's own sentences, arranged. Each label is drawn from a closed vocabulary; the number beside it is the probability the classifier gave that label. No sentence in this record was written by a model.
abstract compiled · 2026-10-15 · no language-model review
How the main claim is stated: asserted 1.00
What the abstract itself states, as probabilities: claims first 0.09 · states limitation 0.10 · human participants 0.05 · controlled comparison 0.17 · randomized assignment 0.01 · states quantitative result 0.07 · compares with prior results 0.34 · mentions code or data availability 0.05
The abstract, sentence by sentence
This table labels the abstract only (arXiv listing abstract v2). The number is the probability the classifier gave the label.
| Sentence | Role | The authors' own work | States the main conclusion | Strength |
|---|---|---|---|---|
| Mathematical reasoning demands two critical, complementary skills: constructing rigorous proofs for true statements and discovering counterexamples that disprove false ones. | background 1.00 | 0.05 | 0.03 | asserted 1.00 |
| However, current AI efforts in mathematics focus almost exclusively on proof construction, often neglecting the equally important task of finding counterexamples. | gap 0.97 | 0.08 | 0.03 | asserted 1.00 |
| In this paper, we address this gap by fine-tuning large language models (LLMs) to reason about and generate counterexamples. | contribution 0.99 | 0.91 | 0.07 | asserted 0.99 |
| We formalize this task as formal counterexample generation, which requires LLMs not only to propose candidate counterexamples but also to produce formal proofs that can be automatically verified in the Lean 4 theorem prover. | contribution 0.62 | 0.65 | 0.07 | asserted 1.00 |
| To enable effective learning, we introduce a symbolic mutation strategy that constructs candidate counterexample-generation tasks by systematically extracting theorems and removing selected hypotheses. | method 0.96 | 0.94 | 0.04 | asserted 1.00 |
| Together with curated datasets, this strategy enables a multi-reward expert iteration framework that substantially enhances both the effectiveness and efficiency of training LLMs for counterexample generation and theorem proving. | method 0.70 | 0.89 | 0.24 | asserted 1.00 |
| Experiments on three newly collected benchmarks validate the advantages of our approach, showing that the mutation strategy and training framework yield substantial performance gains. | result 1.00 | 0.96 | 0.56 | asserted 1.00 |
What kind of paper the classifier reports this is: method or tool 1.00 · evidence from observation or experiment 0.98
What was examined, and what was not
Sources read
- arXiv listing abstract v2
Not examined
- full text
- figures and tables
- references and citing sentences
- relations to other papers
Levels run
- P abstract
Not run
- S sections
- T body sentences
- C citations
- R cross-paper relations
How the labels were made
Compiled by classifier-1.13.0 via TypeSafe API, question bank abstract-p1-20260923, 2026-10-15: 1 request, 5,671 input tokens.
Every quoted sentence on this record is the authors'. Every label is from Pith's closed vocabulary, with the probability shown beside it. This record as JSON
Cited by 2 Pith papers
Reviewed papers on Pith that cite this one. The full citation record
- Honesty Is a Race: Predicting When Verifiable-Reward RL Teaches a Model to Bluff cs.LG · Nov 2026
- Curriculum proof repair with learned counterexamples cs.PL · Nov 2026
How the literature uses it
3 arXiv papers cite this in 3 sentences
Papers citing it · 1 sentence without a confident label
Compares with it 1 paper
-
Formal counterexample generation likewise uses symbolic mutation to construct Lean-checkable negative instances [ref] ; our quality gate instead mutates executable library implementations and asks whether candidate API specifications reject their changed behavior.
P$^{3}$: Joint Program-and-Proof Planning for Verified Code Generation · Specification Quality Pipeline · 0.85
Cites it as support 1 paper
-
Recent work has shown that this capability remains underdeveloped in current LLMs [ref] .
INSPIRE: An Internalize-Then-Improve Approach for Example-Driven Mathematical Reasoning · Introduction · 0.96
Challenges and discussion
Know something about this paper? Leave a comment.
Come back by email
A short list of the week's newly reviewed papers. Nothing is sent until you confirm from your inbox, and every email has a link that stops it. No account.
No comments yet.
Sign in to comment