Add adaptive context compilation study - #4
Merged
Conversation
|
Caution Review failedThe pull request is closed. ℹ️ Recent review info⚙️ Run configurationConfiguration used: defaults Review profile: CHILL Plan: Pro Plus Run ID: 📒 Files selected for processing (30)
🔥 Files not summarized due to errors (1)
📝 WalkthroughWalkthroughThe PR adds experimental SACP-0.2 adaptive context compilation, CLI and JSON-RPC interfaces, adaptive ECEB study support, schemas, replay validation, benchmark results, and documentation. ChangesAdaptive Context and ECEB Study
Possibly related PRs
✨ Finishing Touches📝 Generate docstrings
🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What changed
Why
Fixed context budgets spend the maximum even when an earlier verified packet may be sufficient. This experiment makes expansion explicit and auditable: walk a declared token/node ladder, evaluate a deterministic evidence heuristic, and retain every attempted context digest and stop reason.
The heuristic remains a routing estimate—not a claim of complete evidence or answer quality.
Measured result
On the existing 30 pinned Multi-SWE-bench Rust tasks, the adaptive arm used 4,345 mean prompt tokens versus 5,138 for fixed compilation, retained 51.85% evidence recall, and improved exact fix-path recall from 41.76% to 45.09%. It remained below BM25 and full-context recall. The threshold was calibrated on these same tasks and each arm has one model response per task, so this does not establish quality parity or generalization.
Validation
cargo test --all-features --all-targets --lockedcargo clippy --all-features --all-targets -- -D warningsspec/validate.shscripts/eceb-conformance.shscripts/eceb-study-check.shscripts/eceb-adaptive-study-check.shscripts/site-check.shnpx --yes html-validate site/index.html site/404.htmlcargo package --locked --allow-dirtyThe adaptive replay gate also proves that mismatched selected rounds and premature hard-limit claims fail closed.
Summary by CodeRabbit
adaptive-packcommand andcontext.adaptiveJSON-RPC method, with configurable token, node, coverage, and evidence requirements.