Add Exp 015: combined realistic conditions capstone (#123)#127
Merged
Conversation
Under combined realistic conditions (cacheReliability=0.9, logarithmic growth, tool compression ratio=3, calibrated reasoning), lcm-subagent costs $8.84 at 200 cycles — 2.5% cheaper than incremental. Strategy rankings completely stable across all conditions. The 15-20% combined advantage hypothesis is rejected; reasoning calibration compressed the cost structure, reducing percentage advantages. This is the final experiment in the research programme. All 15 experiments confirm lcm-subagent as the optimal strategy for the Models Agent. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
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
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Summary
Closes #123
Test plan
🤖 Generated with Claude Code