acceptodds
Withdrawn from ICLR 2027

HeurEvo: Agentic Evolution of Hybrid Solver-Augmented Heuristics for Time-Critical Mathematical Optimization

Withdrawn from ICLR 2027, so left out of the paper lists and the map.

Abstract

Recent advances in agentic heuristic design use AI agents and execution feedback to automate algorithm discovery for challenging optimization problems. In many practical settings, high-quality solutions must be obtained under strict runtime constraints, motivating hybrid approaches that combine problem-specific heuristics with powerful mathematical programming solvers. However, existing approaches typically improve heuristic components within predefined procedures or tune solver configurations in isolation. This limits holistic adaptation of where to allocate computation, how to leverage solvers, and how to refine the overall algorithmic structure. To address these limitations, we propose HeurEvo, an *automated plan–code–component co-evolution framework* that jointly evolves the high-level algorithmic structures, their implementations, and a shared pool of reusable components. A planner determines which algorithmic components to use, how to combine them, and how to allocate runtime across stages, a coder realizes the resulting plan as executable code, while a component evolver updates the shared component pool. Within an island-based evolutionary framework, plans and implementations co-evolve with feedback from an interpreter agent that analyzes execution results and identifies opportunities for improvement. Across diverse combinatorial optimization benchmarks, challenging MIPLIB instances, HeurEvo finds high-quality solutions within tight runtime budgets, often matching or surpassing state-of-the-art optimization solvers given hours or days of computation. On several nonlinear geometry problems such as hexagon packing, it also improves upon the best previously reported results. These results highlight the value of jointly searching over algorithmic structure and implementation for agentic heuristic design.

Primary area:
optimization
Keywords:
Automatic Heuristic Design, Agentic Heuristic Design, Combinatorial Optimization, Large Language Models

Reviews

Among:
Shape:
1234this paper, Top 49%
Figure 1. Mean review rating of the reviewed papers at ICLR 2027. This paper averages 2.33 over 3 reviews, above 51% of the others.
Reviewer_rsjMOpenReview ↗
Rating
2
Confidence
4
Informativeness
2
Presentation
3
Soundness
2
Read review
Paper summary

The paper proposes HeurEvo, an agentic evolutionary framework for automatically designing hybrid solver-augmented heuristics under strict runtime constraints. Unlike existing approaches that optimize individual heuristic components or evolve entire programs without explicit structural representations, HeurEvo jointly evolves three levels: algorithmic plans, executable code, and reusable components.

Critical strengths

Novel and practically meaningful co-evolution framework. HeurEvo explicitly separates and jointly evolves algorithmic plans, executable implementations, and reusable components, extending beyond conventional program evolution or fixed-framework heuristic optimization. Its adaptive plan revision and runtime-aware composition address an important practical challenge in time-constrained optimization, making the contribution both methodologically interesting and broadly applicable.

Critical weaknesses

Insufficient accounting for offline evolutionary search costs. Although HeurEvo achieves impressive solution quality under a two-minute execution budget, discovering these heuristics requires substantial offline computation, including numerous LLM calls and program evaluations. The comparison against Gurobi's 48-hour solutions emphasizes inference-time efficiency without fully characterizing the total computational cost or its amortization across problem instances. This weakens the practical efficiency claims and makes it difficult to assess the framework's cost-effectiveness.

Questions for authors
  1. Can the authors quantify the total offline evolutionary search cost, including LLM calls, token consumption, debugging, and program evaluations? How does HeurEvo compare with existing baselines under matched end-to-end computational budgets? This would help establish whether the reported performance gains justify the additional evolutionary complexity.

  2. Are the reported results based on single evolutionary runs or multiple independent runs? Can the authors provide performance variance or confidence intervals across independent evolutionary runs, rather than only across problem instances? This is important for assessing whether the improvements are robust to stochastic search behavior.

  3. For nonlinear geometry problems, how are the extremely small improvements over previously reported best-known solutions validated, particularly when reference values are rounded? Can the authors clarify the feasibility tolerances and numerical verification procedures used to substantiate these claims?

Reviewer_cZrYOpenReview ↗
Rating
3
Confidence
4
Informativeness
1
Presentation
3
Soundness
1
Read review
Paper summary

The paper proposes an agent based algorithm generator where algorithmic components are selected and combined, and updated in the shared component pool after being executed.

Critical strengths

The HeurEvo approach produces competitive results on benchmark problems.

Critical weaknesses

There is limited novelty in the proposed HeurEvo framework. The proposed planner automatically select and combine the reusable algorithmic components, which are implemented by the coder. The only novelty seems to be on the generation of (new?) algorithmic components, updated in the share library. The idea of co-evolution of plans and implementation is not new. It's not clear how the new algorithmic components are assessed before being added to the library. Are there redundant components, and if so are (some and which) of these components are retired?

Questions for authors

See the above comments.

Reviewer_MJgxOpenReview ↗
Rating
2
Confidence
4
Informativeness
3
Presentation
2
Soundness
1
Read review
Paper summary

An automated plan--code--component co-evolution framework was proposed in this paper to solve optimization problems.

Critical strengths

Agentic heuristic design is a good approach to solving optimization problems.

Critical weaknesses
  1. The paper’s novelty could be presented with greater clarity and emphasis. Stating the contributions and hypotheses more explicitly may further improve the manuscript.
  2. The details of the method are not given.
  3. The experimental results are not sufficient to validate the algorithm. The utility of each component is not clear.
Questions for authors
  1. The paper’s novelty could be presented with greater clarity and emphasis. Stating the contributions and hypotheses more explicitly may further improve the manuscript.

From OpenReview, CC BY 4.0. Authors as signed there; shown unedited.

Related papers

Discussion (0)

Sign in to comment.