Continual Learning involves engineering whole systems including Efficient Verifiers to make RL/fine-tuning and running evaluations much chea…

Continual Learning involves engineering whole systems including Efficient Verifiers to make RL/fine-tuning and running evaluations much cheaper at scale! some initial work we’re releasing from LangChain Labs with our awesome friends at Harvey computing reward & correctness in difficult domains like Legal require detailed rubrics that are often driven by LLM as a Judge (instead of programmatically) like we see in coding tasks we find that using open models like DeepSeek V4 Flash and with prompt engineering gets close agreement to using Opus or gpt-5.5 as verifiers while being orders of magnitude cheaper to run some more work coming soon on pushing the Pareto frontier of domain specific verifiers by post-training on Trace data 👀

Source: Harrison Chase (X) | 2026-06-02