REVIEW 5 cited by
Functional Benchmarks for Robust Evaluation of Reasoning Performance, and the Reasoning Gap
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
read the original abstract
We propose a framework for robust evaluation of reasoning capabilities of language models, using functional variants of benchmarks. Models that solve a reasoning test should exhibit no difference in performance over the static version of a problem compared to a snapshot of the functional variant. We have rewritten the relevant fragment of the MATH benchmark into its functional variant MATH(), with functionalization of other benchmarks to follow. When evaluating current state-of-the-art models over snapshots of MATH(), we find a reasoning gap -- the percentage difference between the static and functional accuracies. We find reasoning gaps from 58.35% to 80.31% among the state-of-the-art closed and open weights models that perform well on static benchmarks, with the caveat that the gaps are likely to be smaller with more sophisticated prompting strategies. Here we show that models which anecdotally have good reasoning performance over real-world tasks, have quantifiable lower gaps, motivating the open problem of building "gap 0" models. Code for evaluation and new evaluation datasets, three MATH() snapshots, are publicly available at https://github.com/consequentai/fneval/.
Forward citations
Cited by 5 Pith papers
-
MirrorCode: AI can rebuild entire programs from behavior alone
MirrorCode benchmark shows current AI models achieving up to 56% success reimplementing 25 diverse full programs from behavior alone, including a 16,000-line bioinformatics toolkit.
-
Robust Reasoning Benchmark
The Robust Reasoning Benchmark shows frontier LLMs are mostly resilient to textual perturbations on AIME problems while open-weight models suffer up to 54% accuracy drops and exhibit accuracy decay on later problems d...
-
Probing for Arithmetic Errors in Language Models
Lightweight probes trained on hidden states decode both predicted and correct arithmetic results, detect errors with over 90% accuracy, and transfer from isolated addition to chain-of-thought reasoning.
-
ASyMOB: Algebraic Symbolic Mathematical Operations Benchmark
A new benchmark of perturbed symbolic math problems shows that large language models' performance drops sharply under minor numeric, symbolic, and equivalence transformations.
-
Artificial Intelligence for Mathematical Reasoning: An Integrated Survey of Language Models, Neuro-symbolic Systems, and Verified Discovery
An integrated survey organizing AI mathematical reasoning into informal, formal, discovery, and technique axes while cataloging benchmarks and assessing failure modes.
Discussion (0). Sign in to comment.