- Add csharp case to buildCandidate() with dotnet-script and mcs fallback
- Add defender_strategy.cs.txt seed file (combined Program.cs, Strategy.cs, Grid.cs)
- Add defender seed to population (gamma island, aggression=0.3, economy=0.4)
- Add csharp to langDisplayName() and update comments
Co-Authored-By: Claude <noreply@anthropic.com>
- Updated fitness formula: fitness = 0.7*win_rate + 0.3*kill_rate (was win_rate only)
- Added kill tracking to ArenaResult: TotalKills, TotalMatches, KillRate
- Updated evolver system prompt to explicitly mention combat kills are valuable
- Enhanced arena logging to show kill rate and total kills
This change makes the LLM evolver select for combat aggression, not just win
optimization. The system prompt now informs bots that kills and eliminations
are part of the fitness evaluation, encouraging more aggressive strategies.
Related: bf-59h
- Add NashMixture and MetaWeaknesses fields to meta.Description and
compute them from island population proportions (§10.2 PSRO)
- Update behaviorDistance to support N-D vectors for 4-D MAP-Elites
grid (aggression, economy, exploration, formation)
- Wire NashMixture/MetaWeaknesses through FromMetaDescription converter
so they actually reach the LLM prompt (was dead code before)
- Align LLM prompt with plan §15.1/§15.5: correct combat rules
(focus-fire), fog of war, HTTP protocol section, Nash mixture target
- Fix diversity normalization from sqrt(2) (2-D) to 2.0 (4-D max)
- Rename handleUIFeedback to handleCreateFeedback (§13.6 naming)
- Update tests for new fields and corrected prompt text
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Remove unused encoding/json and net/http imports from cmd/acb-evolver/run.go
that caused build failure. Include other pre-dispatch changes from prior work.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
- Add parent sampling via tournament selection (selector/tournament.go)
- Add replay analyzer to extract key moments, strategies, weaknesses
- Add meta builder for leaderboard summary and dominant strategies
- Add prompt assembler combining parent code + replay + meta context
- Add LLM ensemble with fast tier (GLM-5-Turbo) for bulk generation
and strong tier (GLM-5) for refinement passes
- Add code extraction from LLM responses with language validation
- Add convert utilities for type conversion between packages
- Comprehensive test coverage for all components
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- selector: tournament selection for parent sampling from island populations
- prompt: assembles evolution prompts from parent code, replay analysis, and meta description
- llm: OpenAI-compatible client routing to ZAI proxy with fast (GLM-5-Turbo) and strong (GLM-5) tiers, plus code block extraction from model responses
- Tests for prompt assembly, code extraction, and tournament selection
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
