The Archive

US groups release cheaper models after new challenges to their trillion-dollar ambitions.

We study proactive exploration in LLM agents, i.e., the ability to explore an environment to acquire information that improves future decision-making. In this regard, we first identify two fundamental bottlenecks that hinder this capability and then propose \ours, a novel method designed to instill and refine proactive exploration. Specifically, \ours\ consists of two components: (1) Exploratory Data Construction, which synthesizes exploration-rich trajectories to mitigate the hindsight bias of standard demonstrations; and (2) RL Optimization with Contrastive Signal Guidance, which leverages ...

Reinforcement learning is promising for autonomous urban driving, but long-horizon goal-directed navigation asks a policy to acquire several competing behaviors at once--reaching a distant goal, tracking a route, avoiding obstacles, obeying signals--and a fixed objective gives no order in which to learn them. This paper presents CORAL, which advances two schedules together: a five-stage curriculum that progressively lengthens routes and tightens behavioral constraints, and a stage-aware reward whose component weights shift emphasis from mission progress toward route following, safety, smoothn...

Spatial transcriptomics (ST) enables genome-wide gene expression profiling while preserving tissue architecture, but its cost and limited scalability remain major bottlenecks. This has motivated models that predict spatial expression directly from routine histology. Despite promising results, most existing approaches operate at the gene level without leveraging established transcriptomic modeling practices and rely on heterogeneous gene selection strategies, which complicates fair comparison across methods. We propose to reformulate morphology-to-transcriptomics prediction as conditional gene...

Principle-based regulation, with evaluative standards such as "fair, clear, and not misleading" or "deliver good outcomes", cannot be reduced to binary predicates, and LLM-as-judge is increasingly used as the substitute. Our position is that any such judge must be evaluated on four axes: accuracy, paraphrase robustness, adversarial robustness, and calibration. We release Principle-Bench, 168 cryptoasset financial-promotion scenarios mapped to two UK FCA principles, with paraphrase, adversarial keyword-stuffing, and boundary perturbations authored under a pre-registered rubric; the first bench...

Natural gas prices could triple in some parts of the U.S., which could saddle hyperscalers with massive bills to power their AI data centers.

The anchoring effect is a cognitive bias in which an initial reference value shifts a later judgment toward itself. This effect is well established in human judgment and decision-making, and recent work suggests that large language models (LLMs) exhibit similar behavior. However, existing work on anchoring in LLMs typically evaluates only a narrow set of anchor pathways and rarely distinguishes irrelevant from plausible anchors. We introduce AnchorBench, a benchmark for the anchoring effect in LLMs that evaluates multiple anchor pathways under an explicit anchor relevance axis. Across fourtee...

Meta released Glimmer this week, an open-weight AI model anyone can download and run on their own hardware — a contrast to Muse Spark, the company’s more powerful model that stays locked behind its own APIs. The release landed alongside a letter from Mark Zuckerberg arguing AI should be “for everyone” rather than controlled by a handful of labs, but as Equity’s […]

Apple has reportedly trained a custom AI model for the China market alongside domestic tech giant Alibaba, a rare cross-border partnership that cuts across growing tensions between Beijing and Washington. The China-focused large language model was developed in partnership with Alibaba and trained with the company's support, Reuters reports, citing three unnamed people familiar with the matter. Developing a custom model of its own marks a departure from Apple's previous strategy in the country and would give the company more control over its products in the competitive Chinese smartphone marke...

Congrats to the team!

Down, but not out!

Release: sqlite-utils 4.2.1 Fixes a crashing bug in sqlite-utils 4.2 . I'd introduced code that looks like this: from typing_extensions import Self It turned out the typing-extensions package was not listed as a dependency for sqlite-utils - it was installed by one of the other dependencies in the dev dependency group , but when you uvx sqlite-utils directly you don't get those dependencies. As part of fixing this I figured out how to run a smoke test to ensure the CLI tool still works even without those dev dependencies, which can be run from the project checkout: uv run --isolated --no-defa...

Mico is headed to Learn Live. | Image: Microsoft Microsoft Copilot will no longer show its emotive yellow blob, Mico, when you use the chatbot's voice mode. In a support page, Microsoft says it's going to move Mico to its Learn Live platform, where the avatar will have "more to react to," as reported earlier by GeekWire. Mico launched in Copilot's voice mode last October, with Microsoft AI CEO Mustafa Suleyman pitching the avatar as a way to give the chatbot an "identity." The blob would react to what you say in real-time, complete with facial expressions and animations. Mico joins the long l...

Built as a post-training variation on Z.ai's open source model GLM-5.2, Writer says the new system should provide deployment-ready capabilities at a much lower price.

AI is expensive, Ali Ghodsi tells TechCrunch. With so many investors wanting into his latest round, he said yes to more than planned.

Release: sqlite-utils 4.2 Lots of improvements in this one relating to the table.transform() feature , which adds support for complex alter table operations by creating a fresh table, copying across the data and then dropping and replacing the old one. transform() now preserves a much larger array of edge-case schema definitions, including check constraints, unique constraints and even comments describing the columns. There are also new introspection properties for check constraints, and a whole lot of other smaller changes. Includes contributions from Bunlong Heng , ethanhawkes-gif , Rami Ab...

Release: llm-gemini 0.33 It's been a while since the last llm-gemini release. This version of the plugin adds support for today's Gemini 3.7 Flash release, plus gemini-3.6-flash , gemini-3.5-flash-lite and two embedding models gemini-embedding-2 and gemini-embedding-001 . The plugin is also upgraded for compatibility with LLM 0.32, which means you can now see reasoning traces and you can also enable server-side tools using this pattern: llm -m gemini-3.7-flash -T CodeExecution \ 'use python to calculate (factorial of 13) * 3' I had Gemini 3.7 Flash draw me some pelicans riding bicycles at hig...

Another OpenAI executive is departing. Denise Dresser, who joined OpenAI as its chief revenue officer in December after serving as CEO of Slack, will be leaving in the "coming weeks" to "pursue other opportunities," she said in a team note posted to LinkedIn. Dali Rajic, president and COO of Wiz, will be taking over the CRO role, OpenAI says. Dresser's departure is just the latest in a string of major exits. Earlier this week, special projects lead and former COO Brad Lightcap announced he would be leaving. (Dresser had taken over some of Lightcap's work when he moved to the special projects ...

OpenAI is launching a preview of a sped up version of its latest, most powerful model, in an effort to court enterprise users.

IBM plans to train and certify tens and thousands of consultants on OpenAI's technologies as part of this deal.

Anthropic researchers found AI agents can clash, collude and coordinate in unexpected ways, raising new questions about whether today’s safety tests capture the risks of multi-agent systems.

Transforming multimodal sources into condensed and structured media outputs can be fundamentally conceptualized as a long-horizon agentic process centered on a model-harness system. While an ideal harness system should align with human design priors and accumulate reusable experience through empirical exploration to drive recursive self-improvement, existing paradigms remain static and fall short of this capability. In this paper, we present AutoDesign, a framework that aligns with human design priors, where a meta-harness optimizer guides a code agent to recursively improve harness based on ...

Recent advances in foundation models have enabled AI scientists to automate increasingly complete research workflows, from hypothesis generation and code execution to manuscript preparation. Yet workflow coverage alone does not provide access to the full evidence on which scientific discovery depends. Existing systems typically reason over text, code, labels, or precomputed summaries, leaving scientifically decisive spatial, temporal, cross-channel, and procedural relations unavailable to the agent. We introduce OmniScientist, an end-to-end, omni-modal AI scientist that conducts multidiscipli...

Humanoid motion tracking is central to teleoperation and whole-body imitation, yet evaluation often disagrees with what people perceive in videos. Kinematic errors average per-frame pose differences but miss the physical artifacts that matter most, particularly unstable support and incorrect contacts such as foot skating and mistimed touch-downs. Meanwhile, widely used test suites are small and lack the diversity needed to stress contact-rich, long-horizon behaviors. We introduce HumanTracker to make humanoid tracking evaluation both perceptually aligned and scalable. The HumanTracker benchma...

We study online probabilistic forecasting of binary outcomes chosen by an adaptive adversary. Given an online learning algorithm for a weak hypothesis class $H$, we would like to efficiently obtain two incomparable guarantees that existing online boosting techniques provide separately. Online gradient boosting competes in Brier score with the best predictor induced by the span of $H$ on every sequence, but promises nothing when the span does not contain an accurate predictor. Online weak-to-strong boosting drives classification error to zero under a weak-learning condition, but promises littl...

The per-instance Jaccard score, or intersection over union (IoU), is standard in multi-label classification and binary segmentation. With $s$ labels, its loss matrix has $2^s$ outcomes and reports. Under the convention $\mathrm{Jac}(\varnothing,\varnothing)=1$, we prove that the Jaccard score, shifted-loss, and ordinary loss matrices are nonsingular and that the loss columns have affine dimension $2^s-1$. The proof combines a finite MinHash Gram representation with Boolean Möbius inversion. For exact calibration, we prove $2^{s-1} \leq \mathrm{CCdim}(L^{\mathrm{Jac}}) \leq 2^s-1$. The lower b...

LLM coding agents issue Bash commands through interfaces that may serialize, wrap, and reparse model output. Matched execution scores alone cannot distinguish command-generation errors from failures introduced after generation. QuoteBench measures this boundary with exact final-state validation on 56 one-shot tasks from 14 incident-derived families, crossing the generation contract with the execution transport around one deliberately unescaped added parser. Escaping at the interpolation point reproduces each replayed reply's raw-path outcome, so any recovery under a disclosed boundary must co...

Modern language models are trained on heterogeneous web-scale text corpora. Consequently, studying knowledge and skill acquisition is difficult, as prior exposure to related content is hard to characterize. To address this challenge, we introduce LITTLECURRICULUM, a curated 88B-token pretraining corpus tailored to U.S. elementary school material, explicitly excluding concepts, facts, and vocabulary taught above Grade 5. Training a 5B-parameter LLM from scratch on LITTLECURRICULUM yields LITTLELEARNER, a model with sufficient language competence for open-ended evaluation, yet with clear knowle...

Sparse autoencoders (SAEs) are proposed to extract numerous features from large language model (LLM) representations, yet explaining these features still relies primarily on external observation. This reliance leads to superficial explanations inferred from observed model behavior and computational inefficiency from collecting such behavioral evidence at scale. We introduce SAEVerbalizer, a framework that injects SAE decoder directions into an LLM's representations and fine-tunes the LLM's downstream layers to generate natural-language explanations of the injected features. Once trained, the ...