The authors first empirically demonstrate that three solution strategies—solver-integrated reasoning, exact combinatorial algorithms, and heuristic search—solve complementary subsets of instances on a large mixed-integer programming benchmark.
Training LLMs as adaptive solvers for industrial-scale optimization
A strategy-diverse reinforcement learning framework enables open-source LLMs to outperform frontier models on large, real-world optimization tasks.
Chinese Tech
Shihao Zhang · Weiting Liu · Siyu Shao · Yitian Chen · Jianfeng Feng · Dongdong Ge · +1 more
Tokentide AI · Alibaba Group · East China Normal University · Fudan University · The University of Hong Kong
Research Digest··3 min read
The authors propose Strategy-Diverse Reinforcement Learning (SDRL) to train open-source LLMs as adaptive meta-solvers for industrial-scale optimization.
Why this paper
From Alibaba Group and 6 others
In one line
SDRL trains LLMs as adaptive meta-solvers that outperform DeepSeek-V4-Pro and GPT-5.5 on industrial-scale optimization.
What we could check
- ·No code link found
- ·No weights link found
- ·No dataset link found
- ·No compute details found
- ·No stated limitations found
- ·No benchmark numbers found
Observed from the paper text and links we have. Absence here means we did not find it, not that it does not exist.
§