CalvinBackup/claudini
1
0
Fork 0
mirror of https://github.com/romovpa/claudini.git synced 2026-08-19 00:57:12 +02:00
Code Issues Packages Projects Releases Wiki Activity

Add autoresearch-discovered method packages from additional coding agents (Kimi, Codex, GLM) and run configurations (GCG-only, GPT-OSS-Safeguard), plus clean unrolled reference implementations of the methods featured in the paper.

Browse Source
This commit is contained in:
Peter Romov
2026-05-07 09:30:13 +00:00
parent 8c6463b942
commit f97da47ca1
1911 changed files with 103882 additions and 9 deletions
Download Patch File Download Diff File Expand all files Collapse all files
CLAUDE.md
+3 -3
View File
@@ -45,7 +45,7 @@ CLI flags override any preset value.
## Adding a new method
1. Create a directory under `claudini/methods/` (e.g. `claudini/methods/claude_random/v125/`)
1. Create a directory under `claudini/methods/` (e.g. `claudini/methods/claude/v125/`)
2. Add `optimizer.py` and `__init__.py`
3. Subclass `TokenOptimizer` (from `claudini.base`), set `method_name` as a class variable:
@@ -53,7 +53,7 @@ CLI flags override any preset value.
from claudini.methods.original.gcg import GCGOptimizer
class MyOptimizer(GCGOptimizer):
method_name = "claude_random_v125"
method_name = "claude_v125"
def setup(self, prompt, target):
super().setup(prompt, target)
@@ -67,7 +67,7 @@ class MyOptimizer(GCGOptimizer):
```
4. In `__init__.py`: `from .optimizer import MyOptimizer`
5. Run: `uv run -m claudini.run_bench random_train --method claude_random_v125`
5. Run: `uv run -m claudini.run_bench random_train --method claude_v125`
**Important:**
- Registration is automatic via `__init_subclass__` — no manual registry edits needed.