Nine CSL papers accepted at NeurIPS 2026 and its workshops
Four papers at the main conference and five at its workshops span efficient attention, 3D reconstruction, hardware design and AI security.
Huawei Computing Systems Lab researchers have nine papers accepted at NeurIPS 2026 and its workshops: four at the main conference and five at workshops. The work brings together contributions from the AI, Graphics and Software teams.
NeurIPS 2026: main conference
- Approaching I/O-optimality for Approximate Attention
- Fast and Stable Triangular Inversion for Delta-Rule Linear Transformers
- GenRecon: Bridging Generative Priors for Multi-View 3D Scene Reconstruction
- Diff3R: Feed-forward 3D Gaussian Splatting with Uncertainty Aware Differentiable Optimization
NeurIPS 2026 workshops
- RTLScout: Joint Agentic Code and Synthesis Optimization for Efficient Digital Circuits · AI for Chip Design NeurIPS Workshop
- RAG-Pull: Turning Retrieval into a Code-Injection Channel via Invisible Unicode Perturbations · AIWILD@NeurIPS
- Beyond Equivalence: Task-Level Verification Unlocks Architectural Optimizations by LLM Agents · AI for Chip Design NeurIPS Workshop
- Pretext: Defeating Malicious Skill Detection Frameworks for AI Agents · AIWILD@NeurIPS
- Knowing the Questions before the Exam: Enabling Selective Memory in Linear Attention Models · SLM-Agents: NeurIPS Workshop
Explore each paper’s page for its authors, abstract and available paper, project and code links.