TCSExam: Evaluating Frontier LLMs on the Full Range of TCS Research
Source check: Source checked. Not independent scientific confirmation.
Finding: TCSExam evaluates language-model agents across theorem proving, solver engineering, mathematical constructions and attempts at open conjectures. The 100-task full benchmark and 20-task subset compared eight and nine agents respectively; Lean or executable checkers checked outputs. The benchmark does not establish priority for a mathematical result or capability beyond its sampled tasks.
Technology: Frontier large language models evaluated with Lean and executable checkers
Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.
- AI contribution
- LLM agents attempted proof, solver-engineering, construction and conjecture tasks; checkers validated outputs where applicable.
- Test and result
- Eight agents were evaluated on 100 expert-reviewed tasks and nine on a 20-task subset; Lean kernels or executable checkers validated outputs where applicable.
- Limitation / still unknown
- Benchmark rankings and checked task outputs do not establish priority for any theorem or broad research capability beyond the sampled tasks.
Evidence sources and annotation context
Evidence context (descriptions follow the cited sources; test type is a catalog classification):
Evidence sources: https://arxiv.org/abs/2611.07962v1 · https://arxiv.org/html/2611.07962v1
Source checked · editorial details
The source page was checked against this entry. Independent confirmation is not assessed by this label.
Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.
Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..
No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.
Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.