Metraipolis › archive › dc_ef4737ff3377
AReaL-TIK: Stateful Agentic Optimization of Unified RL Kernels through an Optimization IR — arXiv:2609.35140
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2609.35140
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-07T22:12:15.146Z
- SHA-256 of the archived text
096b33823c4d3ae85c37dea042d55c76086807ffcc58f7571d8529b2c0f56b8c- Stable URL of this page
https://metraipolis.com/documento/dc_ef4737ff3377
How to cite
AReaL-TIK: Stateful Agentic Optimization of Unified RL Kernels through an Optimization IR — arXiv:2609.35140 — arXiv (en), https://arxiv.org/abs/2609.35140 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_ef4737ff3377 (Metraipolis archive dc_ef4737ff3377, sha256 096b33823c4d3ae85c37dea042d55c76086807ffcc58f7571d8529b2c0f56b8c)
BibTeX
@misc{dcef4737ff3377,
title = {AReaL-TIK: Stateful Agentic Optimization of Unified RL Kernels through an Optimization IR — arXiv:2609.35140},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2609.35140},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_ef4737ff3377; sha256 096b33823c4d3ae85c37dea042d55c76086807ffcc58f7571d8529b2c0f56b8c},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.35140. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Reinforcement learning (RL) post-training often uses distinct GPU kernels for rollout and policy update. In synchronous PPO and GRPO, numerical disagreement can perturb ratios between current token probabilities and those assigned during rollout. Recomputing rollout log-probabilities with the policy-update backend avoids this discrepancy but adds a forward pass. Bitwise-consistent unified kernels permit reuse when the policy snapshot and probability processing match the objective. Their optimization must preserve agreement across distinct execution regimes. We present KernelBraid, an agentic framework starting from a hand-tuned, bitwise-consistent implementation. Its optimization intermediate representation (IR) organizes source-code search by linking implementations and modifications to numerical requirements, workload measurements, and derivation history. The agent coordinates changes and retains verified intermediates for further exploration; promotion requires passing correctness checks and improving aggregate latency within per-workload limits. Across 12 end-to-end training configurations on H20, KernelBraid achieves 1.10x average throughput relative to AReaL with log-probability recomputation, and the mean training-reward ratio rounds to 1.00x. Isolated-layer profiling yields 1.40x average speedup in summed phase time across 15 model-GPU pairs. Operator-level evaluation covers correctness and performance for 10 operators on A100, H20, and H200, all passing the prescribed bitwise checks. Unified-attention search achieves 2.52x speedup in summed workload latency over the starting implementation using 7M LLM tokens; ablations assess the contributions of retained evidence and branch exploration to search efficiency and attained performance. Our code is open-sourced at this https URL . Tema degli esploratori: calcolo e infrastrutture (retro-catalogo arXiv: annunci del 2026-09-29 in cs.DC). Categorie arXiv: Distributed, Parallel, and Cluster Computing (cs.DC). Pagina del lavoro: https://arxiv.org/abs/2609.35140 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito). Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina. Letto il 2026-10-07T21:53:13.725Z dalla pagina https://arxiv.org/catchup/cs.DC/2026-09-29?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).