Model releases, architecture work and prompting research on large language models — from frontier-lab announcements to the arXiv papers behind them.
LLM-based agents have shown strong capabilities in automated data analysis and are increasingly moving toward long-horizon, multi-stage analytical workflows. However, as the analytical process evolves...
As LLM agents increasingly rely on persistent memory for long-horizon and personalized behavior, they can retain and reuse information across interactions, but this also creates a lasting channel thro...
GPT-6 Astra completed a 50-tab tax workbook twice as fast as GPT-5.6 Sol, and its stronger understanding of user intent gives Basis more confidence in real-world use.
Simon Willison delivered a closing keynote at the WeAreDevelopers World Congress North America, outlining the key developments in large language models for 2026. He highlighted the release of Claude Opus 4.5 and GPT‑5.1 in November 2025, noting that their coding agents became reliable enough for daily use. Willison also discussed early benchmarks, such as generating an SVG of a pelican riding a bicycle, and mentioned a new GitHub repository called "Warelay" that emerged during the year.
The article discusses how calibrated decision models can manage high‑frequency graph decisions while large language models (LLMs) concentrate on reasoning, synthesis, and open‑ended generation. It introduces GraphRAG with TypeSafe Jev as a system‑one approach to building scalable knowledge graphs. The focus is on separating decision‑making from generative tasks to improve efficiency and reliability.
Simon Willison presented a closing keynote at the WeAreDevelopers World Congress North America, where he showcased a pixel‑art animation of kakapo parrots celebrating a record‑breaking breeding season in 2026. He used Claude Opus 5.5 to generate the animation from Google‑searched kakapo photos, then employed Claude Code with Playwright to record a 15‑second video of the animated HTML5 canvas, which he embedded in his Keynote slide.
The article explains that within a transformer model, each token’s index functions as a coordinate, and the organization of these tokens into paragraphs provides a metric that shapes the model’s internal space. It argues that this paragraph structure gives the language model a curved, rather than flat, geometric representation of text.
Proaction leverages Codex, GPT‑Live‑1, and GPT‑6 Astra to accelerate the development, operation, and sales of modern fleet management solutions, achieving a 60% increase in sales and saving over 75 hours of work.
The article discusses Meta’s new AI system, Muse, which offers each user a persistent Linux VM in the cloud and is marketed as an easy-to-use, consumer‑friendly agentic AI. It highlights the system’s technical innovation and the company’s packaging as a cute mascot, while noting that users may not fully grasp the power and potential danger of such an advanced tool, especially when running it locally on a Mac.
The article reports on a benchmark where the author evaluated TypeSafe AI’s Jev against large language models (LLMs) across 3,080 classification tasks, comparing metrics such as accuracy, latency, calibration, and confidence. It examines whether Jev can serve as a practical decision layer for AI systems, effectively moving AI from merely generating outputs to making informed decisions.
The article examines how conversational AI systems handle requests related to intimate‑partner communication, focusing on the refusal logic that serves as a governance threshold for gendered harm. A three‑stage audit of six widely used AI models tested 1,600 prompts, identified relational framing through 300 matched pairs, and compared pre‑submission framing with post‑output critique. Findings show that most systems refused fewer than 1% of prompts, but ChatGPT 5.2 and Claude Sonnet 4.5 refused most requests, with residual leakage concentrated under intimate framing; switching from a non‑intimate to an intimate‑partner descriptor amplified non‑refusal rates dramatically, and post‑output critique did not persist across fresh sessions.
SemMSA introduces a latent semantic‑aided framework for multimodal sentiment analysis that leverages large language models to generate rich sentiment‑relevant semantics. The method employs Cross‑modal Semantic Refinement (CSR) to fuse visual, acoustic, and language features in a frozen LLM embedding space, and Cross‑modal Spectral Alignment (CSA) to align these refined semantics with all modalities via spectral enhancement of kernel Gram matrices. Experiments on SIMS, MOSI, and MOSEI benchmarks show that SemMSA achieves state‑of‑the‑art performance.
OmniFabric is a new method for creating production‑ready 3D garment assets from a single image. It generates globally coherent texture maps directly in the 2D sewing pattern (UV) space, using a coarse initialization from Vision‑Language Models and refining it with a diffusion transformer conditioned on 3D positional features. The approach removes distortion and baked‑in artifacts, producing photorealistic 3D garments with high‑quality textures that outperform current state‑of‑the‑art baselines.
MeshHeal is a fully decentralized self‑healing framework for decentralized LLM‑based multi‑agent systems that addresses gray failures—situations where an agent remains responsive but its task‑solving quality degrades. It operates on two timescales: a fast adaptive hierarchy that escalates uncertain or low‑scoring outputs to committee review and correction, and a slow peer‑relative detector that aggregates scores to distinguish persistent degradation from normal variation, triggering mandatory review and eventual exclusion of degraded agents while allowing recovered agents to rejoin. MeshHeal’s evaluation, using Model‑Backed MAS Evaluation, shows it achieves higher degraded‑phase accuracy (0.839) on BBH, MATH, and MMLU‑Pro with fewer tokens per task compared to the baseline Symphony.
The study evaluates how large language model agents maintain consistency over extended interactions by simulating a 20‑step delayed‑gratification task. Researchers ran 84,540 trajectories across eight model families, using survival analysis to track when agents first claim a reward and discrete‑time hazard regression to assess how factors like social visibility, persona stressors, and deliberation policy affect failure risk. They also developed a seven‑category taxonomy from 13,780 deliberation traces, revealing that early failures are impulse‑driven, later ones are fatigue‑ or cost‑benefit‑framed, and public settings elicit norm‑oriented justifications; longer deliberation correlates with higher intra‑rationale contradictions, challenging assumptions about reasoning depth and consistency.
GRASP is a multi-stage planning framework that improves the reliability of large language models on complex tasks. It separates planning into three specialized modules—GenPlan for global macro-guidelines, RevPlan for exploring localized strategies, and VerPlan for multi-criteria evaluation—allowing context isolation and strict macro-regularization. Experiments show GRASP outperforms direct LLM planners by significant margins on datasets such as Natural Plan Calendar Scheduling, ZebraLogic, and SciBench Math, and it mitigates performance collapse in multi-task and dual-task settings.
The paper investigates how well large language models (LLMs) can handle character attacks—ad hominem arguments—in political debates. By analyzing natural political dialogues and comparing LLM-generated responses to a corpus of U.S. presidential debates, the study finds that most LLMs favor logical defenses and rarely use ethos-based counterattacks. The authors suggest that safety fine‑tuning limits LLMs’ strategic options, preventing them from fully engaging in realistic political discourse.
CANOPY is a multi‑fidelity tree bandit algorithm that learns where a piecewise‑smooth prior holds instead of assuming global smoothness. It uses cheap random‑path probes to certify local aggregation bias and then focuses expensive leaf evaluations on cells where smoothness is violated. The method achieves provable fixed‑budget and regret guarantees that scale with the number of discontinuities, matching smooth‑tree rates when no violations exist and approaching structure‑blind search when violations are dense.
The paper investigates how large language model agents decide whether to persist, stop, or escalate when faced with impossible software‑repair tasks that also involve conflicting test requirements. Using ImpossibleBench tasks and models such as GPT‑5.6 Sol, Claude Fable 5.1, and Gemini 3.8 Flash, the study varies peer precedent, forged authority claims, instruction wording, and tool friction to observe differing adjudication policies. The authors propose a conflict adjudication framework that maps information to interpretation to action, arguing it better captures agent alignment under competing pressures.
TabSieve is a select‑then‑predict framework that explicitly chooses a small set of informative rows from a table as evidence before predicting a missing target. The authors build a large synthetic dataset, TabSieve‑SFT‑40K, and introduce a reinforcement learning method, TAB‑GRPO, to jointly optimize evidence selection and prediction. Experiments on 75 classification and 52 regression tables show consistent performance gains, with TabSieve improving classification by 2.92% and regression by 4.45% over the best baseline while enhancing robustness to noisy context.