Mechanistic-Interpretability-Study-Induction-Circuit-Stability-Under-Fine-Tuning Mattral · PARTIAL
Tracking exactly what happens to the internal "circuitry" (induction heads) of a 2-layer attention-only Transformer when forced to undergo domain adaptation from prose to structured Python code.
github.com/Mattral/Mechanistic-Interpretability-Study-Induction-Circuit-Stability-Under-Fine-Tuning · ★ 14 · Forks 1 · Size 5.6 MB
SUMMARY
Technologies 13
Scored 10
Observed 10
Practices 6
Evidence 15
Skips 3
COVERAGE
Analyzed 95 files · 159 commits · 0 API calls
TECHNOLOGIES & DEPTH
JSON LANGUAGE Depth 70
1 files · PRODUCTION
Markdown LANGUAGE Depth 70
6 files · PRODUCTION
TOML LANGUAGE Depth 70
1 files · PRODUCTION
YAML LANGUAGE Depth 70
5 files · PRODUCTION
Python LANGUAGE Depth 70
28 files · PRODUCTION
pip BUILD_TOOL Depth 80
1 files · CONFIGURATION
Poetry BUILD_TOOL Depth 80
1 files · CONFIGURATION
Black QUALITY Depth —
0 files · config only
NumPy LIBRARY Depth 81
10 files · PRODUCTION, TEST
RabbitMQ MESSAGE_BROKER Depth 65
1 files · CONFIGURATION
Ruff QUALITY Depth —
0 files · config only
mypy QUALITY Depth —
0 files · config only
pytest TESTING Depth 53
6 files · TEST
PRACTICES
documentation · observedautomated_tests · observedcontinuous_integration · observedcontainerization · absentlinting · observedformatting · absent
ACTIVITY & OWNERSHIP
First commit 2026-03-20
Last commit 2026-06-18
Active months 4
Commits 159