A-CEoH
Algorithmic Contextual EoH
EoH with the A* search code pasted into every prompt, so the LLM designs a guiding heuristic with the algorithm in view.
The loop
- Proposes
Guiding heuristic functions for A* search, as thought plus code
- Judged by
Running A* with the heuristic on instances (solved count, fitness, time)
- Steers
Fitness drives EoH selection; the A* source code is kept in every prompt
Headline result
For unit-load pre-marshalling, Qwen2.5-Coder:32b-generated heuristics solved 30 of 30 held-out instances (seeds 10-39) against 27 of 30 for the "Optimal A*" baseline. For the sliding puzzle they solved 10 of 10 on seeds 0-9 against 5 of 10 for Hybrid A* and for "Optimal A*".
What this does not prove
Two domains. Where both solve, the LLM heuristics give worse solutions than A* (puzzle fitness 0.281 against 0.252 on shared instances). The paper gives no numeric EoH baseline for the final heuristics, only a qualitative claim of improvement. Accepted at EvoStar.
How it generalizes the harness
An evolutionary loop designs a heuristic that is plugged into a classical A* search and then runs without the model.
Classification
| What it proposes | Heuristic |
|---|---|
| Generative model | Pretrained LLMs, frozen (Qwen2.5-Coder:32b, GPT-4o, Gemma2:27b) |
| Algorithmic authority | A* runs on instances (solved or not, plan quality, time) |
| Feedback | Score |
| What gets updated | Population of candidates |
| When it runs | Offline, at design time |
| Validity | Empirical (measured) |
| Evidence | Experiments reported |
| Lineage | Derivative |