Experience Distillation Replication

A faithful replication of Experience-to-Policy Distillation (EPD) for coding agents.

Open source

Experience Distillation Replication media is blocked

Allow external media to connect to the provider and play this content.

Description

This research codebase replicates the Experience-to-Policy Distillation (EPD) method, demonstrating how agents can internalize improvement from prior attempts into a LoRA adapter. It achieves a 25% pass rate on SWE-smith tasks, proving that agents can capture significant in-context learning gains within their weights.

Descriptions, tags, and model credits may be AI-generated or inferred from public sources and can be incomplete or wrong. Learn more.

Details

Why this is interesting: Replicates EPD, proving agents can internalize experience into weights to capture in-context learning gains.
Built with: GPT-5.6 Sol Fable 5 Redwood AI
Links: View tweet
Tags: #coding-agents #distillation #lora #research

Discovered: X
First posted: X · Jul 29, 2026
Listed: Added 3h ago
Build: Strong
Models: Strong