flâneur

Training a State-of-the-Art Legal Agent with Harvey | Applied Compute

appliedcompute.com · 2,693 words · saved by 2 readers

How Applied Compute post-trained GLM-5.1 into the strongest available model on Harvey's Legal Agent Benchmark through full-stack optimization.

We collaborated with Harvey to post-train a frontier legal model on top of GLM-5.1. In Harvey's Legal Agent Benchmark ⌝ (LAB), our trained model outperformed every available model on rubric pass rate. Notably, we found that it outperforms Opus 4.8 Max and GPT-5.5 xhigh, a threshold that previous trained models across the industry had not yet reached. Rubric pass rate All pass eval score GLM-5.1 improves from 0.853 to 0.913 rubric pass rate, exceeding both GPT-5.5 xhigh and Opus 4.8 Max. For all evals, we repeated grading 3 times and reported the average to account for grader variance. We optim

saved by

related reading