The CLRS-Text Algorithmic Reasoning Language Benchmark
This addresses the problem of inconsistent and hard-to-transfer results in LM reasoning research by providing a standardized benchmark, though it is incremental as it adapts an existing benchmark to a textual format.
The authors tackled the lack of standardized benchmarks for evaluating reasoning in language models by introducing CLRS-Text, a textual version of algorithmic traces from the CLRS benchmark, which procedurally generates data for 30 diverse algorithmic tasks and enables fine-tuning and evaluation of LMs as generalist executors.
Eliciting reasoning capabilities from language models (LMs) is a critical direction on the path towards building intelligent systems. Most recent studies dedicated to reasoning focus on out-of-distribution performance on procedurally-generated synthetic benchmarks, bespoke-built to evaluate specific skills only. This trend makes results hard to transfer across publications, slowing down progress. Three years ago, a similar issue was identified and rectified in the field of neural algorithmic reasoning, with the advent of the CLRS benchmark. CLRS is a dataset generator comprising graph execution traces of classical algorithms from the Introduction to Algorithms textbook. Inspired by this, we propose CLRS-Text -- a textual version of these algorithmic traces. Out of the box, CLRS-Text is capable of procedurally generating trace data for thirty diverse, challenging algorithmic tasks across any desirable input distribution, while offering a standard pipeline in which any additional algorithmic tasks may be created in the benchmark. We fine-tune and evaluate various LMs as generalist executors on this benchmark, validating prior work and revealing a novel, interesting challenge for the LM reasoning community. Our code is available at https://github.com/google-deepmind/clrs/tree/master/clrs/_src/clrs_text.