Lumen Research Digest — 2026-03-28

A selective scan of cutting-edge work across AI, automation, graphics, and computer science. This is ranked for novelty and likely significance rather than simply recency.

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. Vega: Learning to Drive with Natural Language Instructions

Source: arXiv
Published: 2026-03-26T17:59:56Z
Why it matters: arXiv: Vega: Learning to Drive with Natural Language Instructions. Vision-language-action models have reshaped autonomous driving to incorporate languages into the decision-making process. However, most existing pipelines only utilize the language modality for scene descriptions or reasoning and lack the flexibility to follow diverse user instructions for personalized driving. To…
Take: Vision-language-action models have reshaped autonomous driving to incorporate languages into the decision-making process. However, most existing pipelines only utilize the language modality for scene descriptions or reasoning and lack the flexibility to follow diverse user instructions for personalized driving. To address this, we first construct a large-scale driving dataset (InstructScene) containing around 100,000 scenes annotated with diverse driving instructions with the corresponding trajectories. We then propose a unified Vision-Language-World-Action model, Vega, for instruction-based generation and planning. We employ the autoregressive paradigm to process visual inputs (vision) and language instructions (language) and the diffusion paradigm to generate future predictions (world modeling) and trajectories (action). We perform joint attention to enable interactions between the…
Link: https://arxiv.org/abs/2603.25741v1
PDF: https://arxiv.org/pdf/2603.25741v1

2. Persistent Robot World Models: Stabilizing Multi-Step Rollouts via Reinforcement Learning

Source: arXiv
Published: 2026-03-26T17:36:08Z
Why it matters: arXiv: Persistent Robot World Models: Stabilizing Multi-Step Rollouts via Reinforcement Learning. Action-conditioned robot world models generate future video frames of the manipulated scene given a robot action sequence, offering a promising alternative for simulating tasks that are difficult to model with traditional physics engines. However, these models are optimized for…
Take: Action-conditioned robot world models generate future video frames of the manipulated scene given a robot action sequence, offering a promising alternative for simulating tasks that are difficult to model with traditional physics engines. However, these models are optimized for short-term prediction and break down when deployed autoregressively: each predicted clip feeds back as context for the next, causing errors to compound and visual quality to rapidly degrade. We address this through the following contributions. First, we introduce a reinforcement learning (RL) post-training scheme that trains the world model on its own autoregressive rollouts rather than on ground-truth histories. We achieve this by adapting a recent contrastive RL objective for diffusion models to our setting and show that its convergence guarantees carry over exactly. Second, we design a training protocol…
Link: https://arxiv.org/abs/2603.25685v1
PDF: https://arxiv.org/pdf/2603.25685v1

3. Back to Basics: Revisiting ASR in the Age of Voice Agents

Source: arXiv
Published: 2026-03-26T17:59:03Z
Why it matters: arXiv: Back to Basics: Revisiting ASR in the Age of Voice Agents. Automatic speech recognition (ASR) systems have achieved near-human accuracy on curated benchmarks, yet still fail in real-world voice agents under conditions that current evaluations do not systematically cover. Without diagnostic tools that isolate specific failure factors, practitioners cannot anticipate which…
Take: Automatic speech recognition (ASR) systems have achieved near-human accuracy on curated benchmarks, yet still fail in real-world voice agents under conditions that current evaluations do not systematically cover. Without diagnostic tools that isolate specific failure factors, practitioners cannot anticipate which conditions, in which languages, will cause what degree of degradation. We introduce WildASR, a multilingual (four-language) diagnostic benchmark sourced entirely from real human speech that factorizes ASR robustness along three axes: environmental degradation, demographic shift, and linguistic diversity. Evaluating seven widely used ASR systems, we find severe and uneven performance degradation, and model robustness does not transfer across languages or conditions. Critically, models often hallucinate plausible but unspoken content under partial or degraded inputs, creating…
Link: https://arxiv.org/abs/2603.25727v1
PDF: https://arxiv.org/pdf/2603.25727v1

4. The Kitchen Loop: User-Spec-Driven Development for a Self-Evolving Codebase

Source: arXiv
Published: 2026-03-26T17:45:00Z
Why it matters: arXiv: The Kitchen Loop: User-Spec-Driven Development for a Self-Evolving Codebase. Code production is now a commodity; the bottleneck is knowing what to build and proving it works. We present the Kitchen Loop, a framework for autonomous, self-evolving software built on a unified trust model: (1) a specification surface enumerating what the product claims to support; (2) 'As a…
Take: Code production is now a commodity; the bottleneck is knowing what to build and proving it works. We present the Kitchen Loop, a framework for autonomous, self-evolving software built on a unified trust model: (1) a specification surface enumerating what the product claims to support; (2) 'As a User x 1000', where an LLM agent exercises that surface as a synthetic power user at 1,000x human cadence; (3) Unbeatable Tests, ground-truth verification the code author cannot fake; and (4) Drift Control, continuous quality measurement with automated pause gates. We validate across two production systems over 285+ iterations, producing 1,094+ merged pull requests with zero regressions detected by the regression oracle (methodology in Section 6.1). We observe emergent properties at scale: multi-iteration self-correction chains, autonomous infrastructure healing, and monotonically improving…
Link: https://arxiv.org/abs/2603.25697v1
PDF: https://arxiv.org/pdf/2603.25697v1

5. Natural-Language Agent Harnesses

Source: arXiv
Published: 2026-03-26T17:58:15Z
Why it matters: arXiv: Natural-Language Agent Harnesses. Agent performance increasingly depends on \emph{harness engineering}, yet harness design is usually buried in controller code and runtime-specific conventions, making it hard to transfer, compare, and study as a scientific object. We ask whether the high-level control logic of an agent harness can instead be externalized as a portable…
Take: Agent performance increasingly depends on \emph{harness engineering}, yet harness design is usually buried in controller code and runtime-specific conventions, making it hard to transfer, compare, and study as a scientific object. We ask whether the high-level control logic of an agent harness can instead be externalized as a portable executable artifact. We introduce \textbf{Natural-Language Agent Harnesses} (NLAHs), which express harness behavior in editable natural language, and \textbf{Intelligent Harness Runtime} (IHR), a shared runtime that executes these harnesses through explicit contracts, durable artifacts, and lightweight adapters. Across coding and computer-use benchmarks, we conduct controlled evaluations of operational viability, module ablation, and code-to-text harness migration.
Link: https://arxiv.org/abs/2603.25723v1
PDF: https://arxiv.org/pdf/2603.25723v1

6. Agent Factories for High Level Synthesis: How Far Can General-Purpose Coding Agents Go in Hardware Optimization?

Source: arXiv
Published: 2026-03-26T17:57:50Z
Why it matters: arXiv: Agent Factories for High Level Synthesis: How Far Can General-Purpose Coding Agents Go in Hardware Optimization?. We present an empirical study of how far general-purpose coding agents -- without hardware-specific training -- can optimize hardware designs from high-level algorithmic specifications. We introduce an agent factory, a two-stage pipeline that constructs and…
Take: We present an empirical study of how far general-purpose coding agents -- without hardware-specific training -- can optimize hardware designs from high-level algorithmic specifications. We introduce an agent factory, a two-stage pipeline that constructs and coordinates multiple autonomous optimization agents. In Stage~1, the pipeline decomposes a design into sub-kernels, independently optimizes each using pragma and code-level transformations, and formulates an Integer Linear Program (ILP) to assemble globally promising configurations under an area constraint. In Stage~2, it launches $N$ expert agents over the top ILP solutions, each exploring cross-function optimizations such as pragma recombination, loop fusion, and memory restructuring that are not captured by sub-kernel decomposition. We evaluate the approach on 12 kernels from HLS-Eval and Rodinia-HLS using Claude Code…
Link: https://arxiv.org/abs/2603.25719v1
PDF: https://arxiv.org/pdf/2603.25719v1

7. Drive My Way: Preference Alignment of Vision-Language-Action Model for Personalized Driving

Source: arXiv
Published: 2026-03-26T17:59:54Z
Why it matters: arXiv: Drive My Way: Preference Alignment of Vision-Language-Action Model for Personalized Driving. Human driving behavior is inherently personal, which is shaped by long-term habits and influenced by short-term intentions. Individuals differ in how they accelerate, brake, merge, yield, and overtake across diverse situations. However, existing end-to-end autonomous driving systems…
Take: Human driving behavior is inherently personal, which is shaped by long-term habits and influenced by short-term intentions. Individuals differ in how they accelerate, brake, merge, yield, and overtake across diverse situations. However, existing end-to-end autonomous driving systems either optimize for generic objectives or rely on fixed driving modes, lacking the ability to adapt to individual preferences or interpret natural language intent. To address this gap, we propose Drive My Way (DMW), a personalized Vision-Language-Action (VLA) driving framework that aligns with users' long-term driving habits and adapts to real-time user instructions. DMW learns a user embedding from our personalized driving dataset collected across multiple real drivers and conditions the policy on this embedding during planning, while natural language instructions provide additional short-term guidance.…
Link: https://arxiv.org/abs/2603.25740v1
PDF: https://arxiv.org/pdf/2603.25740v1

Coverage notes

Candidates considered: 34
Sources included: arXiv plus selected research/lab/blog feeds.
Selection policy: novelty, likely downstream importance, and technical substance.