fix: eliminate 100% temporal drift in CascadingTemporalMemory cold tier#16
Merged
Neal006 merged 1 commit intoMay 24, 2026
Merged
Conversation
…egression When a fact is updated mid-conversation the original value was frozen inside compressed cold summaries, causing temporal_drift_score to report 100% drift at T=100. Two changes fix this: 1. _patch_cold_with_update() — rewrites 'my <key> is <old>' patterns in all cold entries to use the current value whenever an update is detected. 2. _fact_updates dict — stores every seen update so _cascade_warm() re-applies all patches each time a new cold entry is created. 3. merge order flipped: newer cold entry placed first before truncation so current facts survive the 600-char cap. Closes Neal006#2
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
2 participants
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Problem
CascadingTemporalMemoryshowed 100% temporal drift at T=100 (issue #2).Root cause: when a fact's original injection message (
"My city is Bangalore") cascaded from warm → cold before the fact was updated, the stale value was frozen inside the compressed cold summary string.temporal_drift_scorethen countedold_hits=1, new_hits=0and reported drift=1.0.Fix
Three targeted changes to
memory/cascading.py:_parse_update(content)— detects fact-update messages ("my <key> has changed to <new>") and extracts(key_name, new_value)._patch_cold_with_update(cold, key_name, new_value)— rewrites"my <key> is <old>"patterns in all cold summary strings in-place using a regex substitution._fact_updatesdict onCascadingTemporalMemory— accumulates every update seen._cascade_warm()re-applies all patches after each new cold entry is created.add_message()also patches cold immediately when an update arrives, covering the case where the update is absorbed before the next warm→cold cascade.Merge order flip — when two cold entries are merged, the newer entry is placed first so it survives the 600-char truncation limit.
Testing
Closes #2