Fudan University · 2026-04-18 · major
GenericAgent — Fudan's Token-Efficient Self-Evolving LLM Agent With 9k Stars Uses 6× Fewer Tokens
GenericAgent maximises context information density through hierarchical memory, reusable SOPs grown from a 3.3K-line seed, and context compression—achieving full system control with 6× less token consumption than competing agent frameworks.
An agent that grows smarter every time it solves a task
Key specs
| GitHub stars | 9,338 |
|---|---|
| Upvotes | 169 |
What is it?
A general-purpose LLM agent that treats context information density—not context length—as the primary lever for long-horizon performance. Every verified execution path is crystallised into a reusable SOP, building a personal skill tree over time.
How does it work?
Four components work together: minimal atomic tools, hierarchical on-demand memory (only high-level summaries stay in context), a self-evolution mechanism that converts past trajectories into skills, and a compression layer that maintains density during long runs.
Why does it matter?
9.3k GitHub stars within weeks of release. Outperforms leading agents on task completion, tool efficiency, memory, and web browsing while consuming a fraction of the tokens.
Try it
https://github.com/lsdefine/GenericAgent