AI Intelligence Archive

Recent LLMs increasingly adopt hybrid designs that replace standard attention with linear attention, such as Gated DeltaNet (GDN) and Kimi Delta Attention (KDA). Although they compress the context into a fixed-size recurrent state and subs…

The landscape of satellite imagery time series datasets and boundary-pushing architectures for cropland segmentation has never been richer. However, in this gold rush, important truths are being missed on both fronts, as a drive for the mo…

Evaluations of graph reconstruction by language models typically report a single aggregate distance between the original and the reconstructed graph. We prove that for the Wasserstein distance between Laplacian spectra such a summary is br…

Emotion-conditioned text-to-speech (TTS) models may fail to express the requested emotion reliably, and improving controllability by additional training is costly in both computation and emotion-labeled speech training data. We therefore s…

Answering questions about long videos often requires connecting events involving the same objects across hours or days. Chronological descriptions and text-derived entities can leave physical identity unresolved: different objects may shar…

Transformer language models (LMs) are feed-forward: deep-layer representations are never fed back to shallower layers, and the only pathway for information to flow downward across generation steps is the decoded token. This narrow channel …

As agents take on longer and more complex problems, controlling the execution becomes a task in its own right. Each step in the run brings new control choices, like which partial work to build on, whether to start fresh, or when to stop. W…

Agent performance depends on both reasoning ability and the environment in which it acts. We study test-time AI-for-AI, asking how a Builder can learn to construct better execution environments for a Target while both models' weights remai…

A small trainable advisor can steer a frozen language-model executor using natural-language advice. In addition to learning from task rewards, the advisor can use feedback from completed interactions to improve its advice. However, a plaus…

Mixture-of-Experts (MoE), popularized by large language models, is a promising paradigm for scaling visual generative models. However, conventional token-wise MoE routes tokens independently within a homogeneous expert pool and regularizes…

Language-model (LM) harnesses enable LMs to operate effectively over long contexts using additional compute. However, existing long-context evaluations are insufficient for distinguishing modern harnesses, reflected by saturated accuracy a…

Generative modeling is widely used for producing diverse objects from complex, multimodal distributions. However, its expressivity does not, in general, come with formal guarantees that the generated objects satisfy hard constraints or req…

We study average-reward weakly-coupled Markov decision processes (WCMDPs), where a WCMDP consists of $N$ smaller MDPs, called arms, that share multiple per-step budget constraints. We consider the setting where the arms have identical mode…

KV cache memory and bandwidth costs grow with context length and batch size, which limits efficient long-context inference. To address this bottleneck, we introduce WUSH-KV for low-bit KV-cache quantization. It adapts WUSH, which construct…

Verifying a vision-based neural feedback system requires a model of the observations its controller acts upon. Such a model must capture the variation the sensor produces, while remaining tractable for closed-loop analysis. Generative adve…

Many applications of black-box predictive models require controlling task-relevant error rates, such as missed lesion pixels in segmentation or missed labels in multilabel classification. Conformal risk control (CRC; Angelopoulos et al., a…

I threw 629 real AgentDojo attacks at 10 open-source prompt-injection detectors — buried inside ordinary tool output, the way an agent firewall actually sees them. Most are smoke alarms that either sleep through the fire or scream at your …

AI is moving relatively fast, despite being slow down. Yes, it feels like there's a new concept to...

The rapid adoption of artificial intelligence in software development prompts critical questions...

What a refractometer taught me about observability In July, I wrote about a farmer walking a...

I use AI constantly when I code and when I write. I use it to research ideas, plan code, debug,...

I think I've found the simplest way to explain what a prompt can and can't do for safety. Put the...

AI coding tools have changed. A couple of years ago, we were excited when autocomplete correctly...

Vector search is great, but it only goes so far. I benchmarked several ways to improve relevance with surprising results.

AI can generate the implementation. Your architecture still has to decide what that implementation is...

Dear Coder: Open This If You're Feeling AI FOMO Timeless fundamentals beat endless...

Genkit flows live and die with the process that runs them. genkitx-temporal runs any flow as a Temporal Workflow: retries, durable history, timeouts and cancellation.

Building AI apps without a local debugging tool is like writing backend code without a debugger. Genkit Developer UI vs Vercel AI SDK DevTools vs Mastra Studio.

Result I took the AB1-C01 beta on the day it opened and passed with 743 (700 needed). It's...

Genkit Go, Eino, Google ADK Go, tRPC-Agent-Go and LangChainGo compared with real code and real output for each one.

Every LangChainGo concept mapped to its Genkit Go equivalent, with before and after code for models, prompts, chains, tools, agents, RAG and memory.

Ciao 👋 Make some chai, this is a long one, and it has a plot twist I did not see...

I built a synthetic multi-agent loan crew on Amazon Bedrock and made Traccia hard-block a runaway agent, redact applicant PII across every sub-agent, and export EU AI Act audit evidence. Two of my three governance policies blocked nothing …

It didn't make me write better code. It didn't make me deliver features faster. It didn't make the...

This is a submission for the Sanity Challenge, Path Two: Vibe-Code Something Strange What I...

AI agents write this YouTube channel. A satire episode let them take over; here are the real guardrails, plus Amodei's and Bengio's essays on AI agents.

Customer complaints are not the problem definition; they are the signal. This post captures a...

This is a submission for the Kaggle Benchmarking Challenge What happens when an LLM...

A one-line request hides a dozen decisions you never made. TL;DR: Force the AI to ask about the...

AI is everywhere these days. But instead of just learning about AI, why not build something with...