Back to index

Lumen Research Digest — 2026-09-27

A selective scan of cutting-edge work across AI, automation, graphics, and computer science. Previously featured work is excluded. Publications from the last 72 hours come first, with a strict seven-day maximum age.

Big picture

• Agentic and reasoning-heavy systems continue to dominate the high-signal end of AI work.
• Graphics and generative visual research is pushing toward real-time, high-fidelity interactive pipelines.
• Systems work remains tightly coupled to model usefulness through inference, scale, and tooling efficiency.

Selected items

1. OceanXL: Large-scale Underwater 3D Gaussian Splatting via Block Partitioning and Adaptive Pruning

• Source: arXiv
• Published: 2026-09-24T15:35:08+00:00
• Why it matters: Adds new data infrastructure in 3D and visual generation. Stands out for unusually strong scope and useful downstream control.
• Summary: We propose OceanXL, a fast and scalable 3DGS-based framework for large-scale underwater reconstruction. We also introduce a large-scale underwater dataset covering diverse marine environments. OceanXL is best read as new data infrastructure in 3D and visual generation.
• Link: https://arxiv.org/abs/2609.29985
• PDF: https://arxiv.org/pdf/2609.29985

2. Jev-Mobile: Jev as an Executor for Mobile GUI Agents

• Source: arXiv
• Published: 2026-09-24T17:30:32+00:00
• Why it matters: Adds an implementation framework in systems efficiency. Stands out for for operational use cases.
• Summary: We introduce Jev-Mobile, which shifts this paradigm to low-frequency VLM planning and high-frequency lightweight execution: the VLM specifies local goals, the accessibility tree defines a structured executable action space, and Jev, a fast typed decision…. These results show that decoupling high-level VLM reasoning from low-level action execution can substantially improve mobile GUI agent efficiency while maintaining competitive task performance. Jev-Mobile is best read as an implementation framework in systems efficiency.
• Link: https://arxiv.org/abs/2609.30186
• PDF: https://arxiv.org/pdf/2609.30186

3. Industrial Anomaly Detection via Defect-Grounded Reasoning in Visual Latent Space

• Source: arXiv
• Published: 2026-09-24T12:11:18+00:00
• Why it matters: Adds new data infrastructure in agent debugging and observability. Stands out for unusually strong scope.
• Summary: To address these, we propose Anomaly-LR, a defect-grounded latent reasoning framework that first forms a global understanding of the input and then progressively refines anomaly-relevant representations directly in the visual latent space. We further construct IAD-LR-22K, the first IAD instruction dataset designed for latent reasoning, containing 22,228 image-question instances from 4,523 industrial images, with global textual reasoning traces and region-level visual annotations. Industrial Anomaly Detection via Defect-Grounded is best read as new data infrastructure in agent debugging and observability.
• Link: https://arxiv.org/abs/2609.29457
• PDF: https://arxiv.org/pdf/2609.29457

4. AdaHVLA: Adaptive Harnesses for Long-Horizon Vision-Language-Action Execution

• Source: arXiv
• Published: 2026-09-24T08:14:03+00:00
• Why it matters: Adds better debugging hooks in robotics and embodied perception.
• Summary: To bring these complementary capabilities together, we introduce AdaHVLA, an adaptive harness that refines code-based coordination policies through robot experience to better align agent reasoning and memory with VLA execution. Title: AdaHVLA: Adaptive Harnesses for Long-Horizon Vision-Language-Action Execution Base summary: Vision-language-action (VLA) models offer strong local control and instruction following but often struggle with long-horizon tasks requiring persistent memory…. AdaHVLA is best read as better debugging hooks in robotics and embodied perception.
• Link: https://arxiv.org/abs/2609.29204
• PDF: https://arxiv.org/pdf/2609.29204

5. When Can Agents Forget Their Reasoning? ICLR for Long-Horizon Agent Context Compression

• Source: arXiv
• Published: 2026-09-24T14:28:43+00:00
• Why it matters: Adds better debugging hooks in agent workflows.
• Summary: We propose Interaction Aware Compression for Long Horizon Reasoning (ICLR), a training free online method that ranks reasoning blocks using frozen proxy entropy while preserving actions, tool calls, and observations. ICLR for Long-Horizon Agent Context Compression Base summary: Long horizon language model agents continually accumulate reasoning history, increasing context length and inference cost even after earlier decisions have been executed and observed. When Can Agents Forget Reasoning is best read as better debugging hooks in agent workflows.
• Link: https://arxiv.org/abs/2609.29875
• PDF: https://arxiv.org/pdf/2609.29875

6. PolyUMI: Accessible Visual-Tactile-Audio Data Collection for Object Inference and Manipulation

• Source: arXiv
• Published: 2026-09-24T13:09:50+00:00
• Why it matters: Adds an implementation framework in multimodal perception.
• Summary: Experiments spanning object inference, slip control, and contact-rich manipulation show that touch and audio reveal task-relevant information beyond vision and that VisTA is competitive with or outperforms existing multimodal policies. To effectively use these heterogeneous observations, we further introduce VisTA, a token-level multimodal policy that integrates information across sensors and time to predict contact-aware robot actions. PolyUMI is best read as an implementation framework in multimodal perception.
• Link: https://arxiv.org/abs/2609.29760
• PDF: https://arxiv.org/pdf/2609.29760

7. SARFusion: Scene-Aware Routing Fusion for Robust Camera-LiDAR 3D Object Detection

• Source: arXiv
• Published: 2026-09-24T08:43:04+00:00
• Why it matters: Adds better debugging hooks in 3D and visual generation.
• Summary: However, existing fusion detectors often establish strong inter-modality dependencies by decoding object queries from tightly coupled multimodal representations. To bridge this gap, we reformulate robust camera-LiDAR fusion as a scene-aware branch routing problem and propose SARFusion, a robust 3D object detector. SARFusion is best read as better debugging hooks in 3D and visual generation.
• Link: https://arxiv.org/abs/2609.29235
• PDF: https://arxiv.org/pdf/2609.29235

Coverage notes

• Candidates considered: 616
• Sources: scientific papers from official arXiv new-paper announcements, with the arXiv API as fallback. Published dates are original submissions verified on official arXiv abstract pages, not announcement or revision dates. Revisions and company news are excluded.
• Selection policy: never repeat featured work; prefer the last 72 hours; exclude publications older than seven days or with unknown dates. Fewer qualifying items means a shorter digest.