AI Research Pulse

The case rests on a practical change: LLM agents no longer need the tagging that stalled Berners-Lee's semantic web. The author argues we are facing...

This paper tackles edge surveillance's compute and privacy costs by working on pose landmarks instead of video.

• Privacy by design: The app does not...

The authors frame the goal as a voice assistant that can "understand requests, reason over context, sustain conversations, and act through tools"....

Two papers target the same weakness in World Action Models from different angles: transferring to new tasks, and identifying what actually drives...

This paper argues that most adaptive math tools rely on static or rule-based personalization, and that conventional approaches do not effectively...

This IJCAI 2026 main-track paper is keyworded under planning and scheduling and targets the 3D bin packing problem, which is widely applied in...

UPR Mayagüez has secured an $8 million NSF grant to establish CAPCell, a center combining AI, cellular manufacturing and bioprocess engineering.

-...

The paper targets a gap: existing methods model short-term dynamics and long-term objectives separately, and feed predicted futures into action...

Robot control harnesses, the code wrapped around a frozen policy, are typically hand-built and redesigned for each new robot or benchmark....

Multi-agent systems need both effective coordination and specialized task execution, yet existing prompt optimization typically couples these...

The recipe used to train language-model agents, supervised fine-tuning followed by multi-turn RL, is being applied to robot manipulation through...

Proactive agents must decide whether, when, and how far to help without being asked. Automated scores and binary verdicts offer limited insight into...

A study embedded a consistent recurrent neural network (ConsRNN) surrogate as a fracture criterion in Abaqus/Explicit shell simulations. The available excerpt is brief, so the details of the implementation and its validation are not visible here.

A new paper presents a spiking neural network demodulator for event-camera optical camera communication that works with pulse-shaped waveforms. The...

Leaderboard scores work best at the top of the funnel. A practical stack starts with your own tests.

• Start with tasks you can read: The guide...

The authors ask whether AGI progress could come from pairing a lightweight System-1 predictor with System-2, rather than scaling System-2 alone. Their...

Meta reversed the usual robotics playbook. According to reporting, it started with Muse, its consumer AI agent, and then set out to build a body for...

An AI model can perform well in a test environment yet struggle in real use. Different data, changing conditions, or an unexpected use case can expose...

On-policy RL for LLMs assumes training responses come from the current policy. In practice, an inference engine samples responses while a separate...

Start with end-to-end numbers before you optimize:

• Tool: The guide recommends trtexec for benchmarking ONNX models with TensorRT. It builds a plan...