Method: Memory-Enhanced Designer Boosts Agents by 10% at 30B Scale
natashajaques · x · 2026-08-20
This approach improves the Reasoning Agent by addressing the bottleneck in traditional Designer methods, which often lack task diversity. By giving the Designer memory and conditioning it on samples from a large pre-training corpus every episode, task diversity is significantly enhanced.
Results show:
- Performance improves with model scale.
- At the 30B scale, the method achieves about a 10% improvement over the base model on benchmarks like LiveCodeBench-v2, tau^2-bench, and BFCL.
- The generated environments are diverse, interesting, and increase in complexity throughout training, offering an alternative to manually designing RL training environments.
Related event: Boosting Agent Training: Memory for Task Designers Yields 10% Gain(2 posts)→
More from coding & agent
- Meta open-sources Muse Spark Open agent model with day-0 TPU support in SGLang — BanghuaZ · 2026-10-16
- Saying "load-bearing" in your prompt makes Opus 5.5 stop holding back — zacharynado · 2026-10-16
- You.com partners with Paperclip to give every agent live web access — PolarBearby · 2026-10-16
- I priced every Grok Bot call: $20 to $200 plan in 5 days, long jobs eat 52% — jonathanmalkin · 2026-10-16
- GitHub is criminally underrated as a free data store, devs argue — blelbach · 2026-10-16
- Claude Desktop doesn't support MCP sampling — here's a workaround for two-way MCP apps — wuj · 2026-10-16