{"slug": "optimize-the-cob-then-the-kernel-optimizing-beam-search-for-generative-inference", "title": "Optimize the Cob, Then the Kernel: Optimizing Beam Search for Generative Recommendation Inference", "summary": "A team of researchers reports in arXiv paper 2611.23339v1 that optimizing the generation pipeline holistically, without custom kernels, cuts end-to-end latency for generative recommendation inference by up to 25x versus vLLM on Ascend NPUs. The approach concatenates beams along the sequence dimension using Tree Attention to eliminate redundant KV cache accesses over the shared prompt prefix, and adds pre-generated partial-attention masks, static KV cache allocation, loop-unrolled generation dispatch, TorchAir graph compilation, automatic operator fusion, and vocabulary-constrained decoding. The authors argue the generation pipeline should be optimized holistically before resorting to custom kernel development, noting their generation latency is comparable to a state-of-the-art solution that uses custom kernels.", "body_md": "arXiv:2611.23339v1 Announce Type: new \nAbstract: Generative recommendation (GR) inference differs substantially from conventional LLM serving, featuring long input sequences, wide beam search, and only a few fixed output tokens. Existing LLM serving engines such as vLLM process beams as batched independent sequences, causing redundant KV cache accesses over the shared prompt prefix. Furthermore, vLLM executes each beam search generation step as a separate engine invocation, incurring host-side beam processing overhead between decoding steps and substantial runtime KV block management in PagedAttention. Using Tree Attention, we concatenate beams along the sequence dimension to eliminate redundant KV accesses, and further optimize execution through pre-generated partial-attention masks, static KV cache allocation, loop-unrolled generation dispatch, TorchAir graph compilation, automatic operator fusion, and vocabulary-constrained decoding. Without the use of custom kernels, our approach reduces end-to-end latency by up to 25x over vLLM on Ascend NPUs, and achieves generation latency comparable to a SOTA solution using custom kernels, showing that the generation pipeline should be optimized holistically before resorting to custom kernel development.", "url": "https://wpnews.pro/news/optimize-the-cob-then-the-kernel-optimizing-beam-search-for-generative-inference", "canonical_source": "https://arxiv.org/abs/2611.23339", "published_at": "2026-11-19 05:00:00+00:00", "updated_at": "2026-11-19 05:12:45.599517+00:00", "lang": "en", "topics": ["ai-research", "ai-infrastructure", "mlops", "large-language-models"], "entities": ["vLLM", "Ascend NPUs", "TorchAir", "PagedAttention", "Tree Attention", "arXiv"], "also_reported_by": [], "takeaway": "Concatenating beams along the sequence dimension with Tree Attention plus static KV cache allocation, loop-unrolled dispatch and TorchAir graph compilation cut generative-recommendation inference latency up to 25x versus vLLM on Ascend NPUs without custom kernels.", "recommendation": {"advice": "For beam-search generative recommendation on Ascend NPUs, batch beams via Tree Attention with pre-generated partial-attention masks and static KV cache before writing custom kernels; the source measured up to 25x lower latency than vLLM.", "playbook": "inference-serving", "playbook_url": "https://wpnews.pro/playbooks/inference-serving"}, "alternates": {"html": "https://wpnews.pro/news/optimize-the-cob-then-the-kernel-optimizing-beam-search-for-generative-inference", "markdown": "https://wpnews.pro/news/optimize-the-cob-then-the-kernel-optimizing-beam-search-for-generative-inference.md", "text": "https://wpnews.pro/news/optimize-the-cob-then-the-kernel-optimizing-beam-search-for-generative-inference.txt", "jsonld": "https://wpnews.pro/news/optimize-the-cob-then-the-kernel-optimizing-beam-search-for-generative-inference.jsonld"}}