Repository navigation
[codex] Export public decision settlement feedback - #8
Merged
Merged
Conversation
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Motivation
Round-level reward research needs observed unit contributions and attributed economic outcomes after decisions. The existing decision ledger exposes public progress, while battle counters and economic provenance remain inside Core.
Changes
Coverage and limitations
Economic coverage is explicitly partial. investShip non-currency outcomes and other economic sources remain missing. Nested invocation deltas may include milestone payments that also have individual payment events, so downstream consumers need deduplication. Direct unit kills use deployment.kills and have a different scope from spawn-ledger creditedKills. Blocking count and duration remain unavailable. Counter retention completeness and implemented content coverage are separate concepts.
The feedback lives outside observation and policy encoding. This PR supplies measured feedback for subsequent diagnostics and research reward design.
Compilation
Both compilation checks completed locally. The existing native-build pull-request workflow builds Linux, Windows and both macOS targets on Python 3.12.