Evolving Initial Heuristic Functions for Agent-Centered Heuristic Search
Why this work is in the frame
A frame that forgets how it found something cannot be audited. These are the routes that admitted this work.
Bibliographic record
Abstract
Heuristic functions guide search algorithms and have a profound impact on their performance. In the context of agent-centered real-time heuristic search (RTHS), a heuristic represents the agent's initial domain knowledge which the agent then updates as it explores the search graph. An ideal initial heuristic should capture some specific domain knowledge to guide the agent effectively yet be general enough for a broad class of problems. It should also be computationally efficient, compact in its representation and human-interpretable. Traditionally initial heuristics in RTHS have been designed by humans (e.g., Manhattan distance). In this paper we explore the alternative of building initial heuristics by machines. To keep them portable and human-interpretable we represent each heuristic as a closed-form algebraic formula. Yet to make the heuristics capture problem specifics and thus be more effective in guiding the search, we automatically build a heuristic tailored to a class of problems. To achieve both objectives, we propose and evaluate automatically searching the space of heuristic functions. As a preliminary demonstration, we find closed-form heuristics that outperform Manhattan distance in grid-based pathfinding. We then develop an insight on how such formula-based heuristics are able to exploit characteristics of certain pathfinding maps.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.000 | 0.001 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.000 |
| Open science | 0.001 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.001 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from it