The Archive

Roadside camera-radar 3D object detection adapts end-to-end models for autonomous infrastructure despite varying sensor mounting geometry.

When Nvidia originally showed off DLSS 5 earlier this year, the technology immediately got branded as gen AI for video games. To gamers, Nvidia had lost the plot. I personally couldn't get over how faces were transformed using DLSS 5. The 15-year-old student from Hogwarts Legacy suddenly looked much older, and Grace from Resident Evil Requiem looked yassified. Now more than six months later, DLSS 5 is about to debut properly in Hogwarts Legacy, and it looks like it has been seriously toned down. Nvidia published new comparison screenshots today, where the changes to faces are as subtle as the...

Factuality alignment methods lack independent reproduction and fair comparison across base models; reproducibility study surfaces methodological gaps.

Unigram surprisal estimation in psycholinguistics breaks probabilistic link between contextual and marginal distributions; proposes principled alternatives.

Prompt optimization, prompt tuning, and prefix tuning alter dense/MoE models differently; task information flow depends on adaptation location.

Graph-modified embeddings offline enable multi-hop retrieval without query-time graph access; outperform direct graph replacement via multi-stage strategies.

Linear probes on a model's hidden states are used as cheap monitors for indirect prompt injection: they screen the tool outputs an agent reads and send suspicious ones to a slower second check, so every false alarm has a cost. But a probe learns whatever separates its training examples. The standard evaluation of injection probes can be passed without the model: conversation position and text lengths alone reach 0.91-0.98 AUROC. When each attack is compared with the same conversation without the injection, or with harmless text formatted like an injection, standard probes flag most of these h...

Diffusion language models reconstruct masked tokens in parallel, offering an alternative to autoregressive generation. Yet their Transformer backbones typically carry visible evidence and evolving masked predictions through a single residual stream. We equip masked diffusion models with multiple residual streams that persist through depth and adaptive routing between them, then investigate how they support these complementary roles. On the pretraining task, two streams reduce reconstruction perplexity by 5.71% on validation data and 5.13% on average across external corpora compared with stand...

The new X Subscription combines X's paid features with AI tools like Grok and Cursor, offering plans ranging from $8 to $200 per month.

First large-scale continual learning benchmark for radar waveform classification, addressing catastrophic forgetting in deployed classifiers.

LunaTP accelerates Clebsch-Gordan tensor product convolutions in equivariant MLIPs for molecular simulations via node-centric parallelism.

FEDE framework quantifies predictive causality by measuring forecast-error entropy changes across prediction horizons in nonlinear time series.

PRISM uses counterfactual probes to calibrate LLM judges, correcting implicit scoring biases by measuring sensitivity to attribute emphasis.

General calibration harness (GCH) enables reproducible delegation of Earth science model calibration to knowledge-guided LLM agents.

Geoffrey Huntley describes using pre-commit hooks to enforce agent compliance with internal diagnostics libraries, preventing drift in instrumentation.

Theory of representation reuse defines task-relevant state as coarsest deterministic state sufficient for families of downstream tasks.

Attack breaks multi-bit LLM watermarks without key, remapping watermark IDs by rewriting text while preserving meaning.

Urbex agentic framework grounds landmarks in city-scale 3D scenes via active spatial search and RL-optimized tool-use policies.

ABot-ER unified embodied reasoning model achieves specialist depth across perception, planning, execution via data-model co-evolution.

Strategic LLM selection extends game-theoretic benchmarks to model routing, choosing LLMs considering opponent adaptation in multi-agent settings.

The HomeView is a smart speaker with a touchscreen display. It’s available in silver, starlight, rose pink, and space grey. | Image: Apple It's finally here: Apple's long-rumored smart display, the Apple HomeView, has arrived. I got a first look at the device in New York City this week, and it's gorgeous. The HomeView puts a small, vibrant screen atop a speaker base for taking video calls, beautifully displaying photos, and providing information at a glance with Apple's Siri AI voice assistant. The device's square shape makes it feel less imposing than larger smart displays like Google's Nest...

American Express is introducing a new framework designed to ensure AI agents only make purchases their users have authorized, potentially preventing costly mistakes and disputes as AI-powered shopping grows.

The startup will sell some of its assets to Disney for $20 million.

OpenAI's Higgsfield enables video generation beyond 10 minutes using GPT-6 Astra, GPT-Image-2.5, and GPT-Live for voice-driven filmmaking.

Character.AI, one of the first breakout startups of the generative AI boom, has filed for bankruptcy and is selling some of its assets to Disney. The company, which builds various online bots for users to chat or roleplay with, has faced mounting scrutiny and lawsuits over the safety of its chatbots, but has continued releasing updates and products in recent months. In a press release and blog post published Monday, Character said it had filed for Chapter 11 bankruptcy protection in the US to facilitate the sale of the company and its assets. "Over the past year, we've taken meaningful steps ...

SpaceX acquires additional spectrum for mobile strategy, with Stratechery analyzing market structure implications.

MIT Technology Review Explains: Let our writers untangle the complex, messy world of science and technology to help you understand what’s coming next. You can read more from the series here. Late last month, Anthropic released a report arguing that the model GLM-5.3 poses a major cyberattack risk. The report’s core concern is that the…

MSL’s emergence as a frontier lab is finally complete.

When Keelan Miskell began winding down his crypto startup this year, he didn’t think its data was worth much: the users were anonymous, the company just a few years old. But he’s since turned the firm’s digital archives into real money. He’s netted roughly $25,000 for his code base and a still-pending amount (likely around…

web-edit 0.1a1 alpha release includes performance optimizations generated by Claude Opus 5.5 from profiling data.