” Full replay serves all boundaries from the recording; cut-point replay serves only a selected subset and executes the remaining components live using changed code.
Cut-point replay turns recorded agent failures into regression tests
Chronicle selectively replays recorded model and tool interactions while running changed components live, allowing past failures to become continuous-integration tests.
Big Tech
Tisha Chawla · Susheem Koul
Microsoft
Research Digest··2 min read
Thread:Agent Self-Improvement
Chawla and Koul built Chronicle, a record-and-replay system that captures an LLM agent’s non-deterministic boundaries as immutable records.
Why this paper
From Microsoft · Released code · Part of Agent Self-Improvement, now 11 papers
In one line
Cut-point replay converts recorded LLM agent failures into deterministic regression tests that detect faulty code changes.
What it released
Code
What we could check
- ✓Code link in the paper (github.com)
- ·No weights link found
- ·No dataset link found
- ·No compute details found
- ✓Limitations stated by the authors
- ·No benchmark numbers found
Observed from the paper text and links we have. Absence here means we did not find it, not that it does not exist.
§