You Don’t Need Astra to Edit a Model’s Weights
—A GPT-OSS follow-up to Weight-Space Programming: direct edits to a small LunarLander policy, with held-out evaluation.
5 writings
A GPT-OSS follow-up to Weight-Space Programming: direct edits to a small LunarLander policy, with held-out evaluation.
Tool use in hidden, stateful environments.
Experiments in sparse feedback and agent memory.
Training large models within memory limits.
One embedding table for reading and writing tokens.