TBNAI
Pareto-Optimal RTL Code Generation via Multi-Objective Reinforcement Learning with Large Language Models
Publication: cs.AI updates on arXiv.org
Author: Koki Takeshita, Kenji Uebayashi, Hikaru Tomonari, Masaya Tsunokake, Koichiro Yawata, Tatsuya Sasaki, Yusuke Ohtsubo, Kota Dohi, Yasuhiro Sogawa
Source published: 21/10/2026, 04:00:00 UTC
Briefing created: 21/10/2026, 04:16:40 UTC
Based on a feed excerpt; full article content has not been acquired. Source evidence retrieved 21/10/2026, 04:03:24 UTC.
Briefings
This research introduces a new method for LLM-based RTL code generation that optimizes the diversity of the entire Pareto-optimal solution set rather than just a single circuit. Evaluated on the RTLLM v2 benchmark, the approach achieved +13.3 percentage points in relative Hypervolume compared to a baseline without PPA optimization and +14.4 points over a single-objective RL method, demonstrating superior circuit diversity.
Key points
- The study proposes a learning method for LLM-based circuit generation that optimizes the quality and diversity of the entire Pareto solution set through multiple generations.
- The method uses a reward function based on Hypervolume Contribution (HVC) to measure how much each generated circuit contributes to the Pareto front.
Context
Recent studies propose reinforcement learning with verifiable rewards (RLVR) to generate high-quality RTL circuits from natural-language specifications. Existing approaches optimize a single circuit's PPA (Power, Performance, Area) by aggregating objectives, making it difficult to generate diverse trade-offs. This work is the first to optimize the quality and diversity of the entire circuit set obtained through multiple generations.
Limitations
Evaluation was conducted on the RTLLM v2 benchmark. The study compares the proposed method against a baseline without PPA optimization and a single-objective reinforcement learning method, but does not provide results for other specific benchmarks or compare against other RLVR methods.
Generated by Bittrees-hosted qwen35-2b sha256:aaf42c8b7c3c on 21/10/2026, 04:16:40 UTC. Based on publisher-provided feed evidence; may not cover the full original article. Check the source for details.
View archived briefing on IPFS
bafybeiefcnkxvesmqtfl7qx6mxf7uvjyws2pm6iljqt5nzf7i7l3taajmq