A preprint is a signal, not a finished product or an independently confirmed result. We therefore track research papers separately and show the actual state of supporting evidence for each one.
1381
published research events
Latest work
A significant claim from a single source is published only after further confirmation.
A review describes the use of deep learning and generative models to detect defects in nondestructive testing. It identifies a lack of standardized open datasets and difficulties with model explainability as barriers to adoption.
GHARP separates the preparation of human geometry and appearance from real-time facial expression animation. According to the authors, it achieves state-of-the-art quality on Ava-256 and runs up to 13× faster on an A100 GPU while using 8× fewer Gaussians.
OpenAI presents model results concerning the Kakeya conjecture, Exponential Time Hypothesis and quantum chromodynamics. The available excerpt mentions ways for researchers to verify the results, but does not provide specific findings or a method.
Researchers connected an LLM to a classical simulator to iteratively design and optimize quantum circuits. According to the study described, across eight tasks the system reduced the combined requirements for qubit count and computation time by approximately 4× to more than 9 100× compared with reference designs.
Part of a broader thematic overview; only this AI event has been covered.
P-Zero Research introduced the TasteVal benchmark with eight research tasks and a comparison group of 24 experts. According to the reported results, the best-performing model tested outperformed the human baseline. However, the test does not assess autonomous AI self-improvement.
In a study by Convovo, the Grok 4.7 model recommended Republicans in 92 % of responses, and the GPT-6.1 Sol model recommended Democrats in 97 %. The test covered nearly 500 electoral contests and forced a choice of candidate via the API. Consumer apps refused to answer when given the same prompt.
The authors compared Claude Code, OpenCode and mini-SWE-agent using the same models. Across 447 tasks, Claude Code and mini-SWE-agent were less than two percentage points apart for both tested models. According to the study, costs can rise up to threefold.
At IDWeek 2026, the Google Research team summarizes the development of AI for tuberculosis screening using X-rays. It refers to a comparison with radiologists from 2019, an independent evaluation from 2021 and clinical validation from 2024. The available excerpt does not describe a new release.
The authors presented LitPivot, a system that links revisions to research proposals with literature searches and evidence-based critique. In a study with 17 researchers, experts rated the resulting ideas as better grounded in the literature than those developed using a text editor and AI chat.
According to Anthropic, Claude generated about 180 million lines of proofs for 123 libraries over several weeks and uncovered hundreds of bugs. The proofs are checked by Rocq; they apply only to a defined architecture, single-threaded programs and inputs under 64 MB.
The study formalizes permission, approval and budget checks when LLM agents call tools. It derives the safety of the running system under four assumptions and shows that repeated planning can allow stricter controls to result in fewer interventions.
Latent Interconnect Protocol connects six specialized models to an LLM using trained connectors; the original models remain unchanged. The authors report transfer to tasks across domains without joint training, but warn that input interpretation is unreliable.
The authors present SARL, a reinforcement learning method for recognizing the state of a search without additional evaluation at inference time. They report a relative improvement of 24.3 % over the strongest baseline; the available abstract does not specify the metric.
The RoleFlow system adjusts the structure of collaboration among agents and creates roles as needed. The SECPO method compares each adjustment with keeping the current state unchanged. The authors report a lead of 4.7 points over the strongest competitor on average across seven test suites.
The authors introduced SGUR, a method that requires no additional training. In a diagnostic test derived from PrimeKG, they report that it achieved 92.5 % Hit@3, 9.4 percentage points higher than the strongest comparison method, Graph-RAG, while using approximately one-fifth as many tokens.
The study introduces FusionEval with eight public benchmarks. According to the authors, methods that require no additional training preserve reasoning ability but significantly impair tool use. The MOPD method strikes a better balance between these capabilities, but requires additional data preparation and training.
The authors introduce minimal-edit on-policy distillation: they make targeted corrections to incorrect model outputs during training. According to the abstract, the method improves geometric perception and does not require reference image descriptions when the model is used.
The TRIAD benchmark contains 327 questions in 86 groups of scenes and tests the combined use of text, images and audio. According to the study, the best of the 14 systems evaluated trails humans by more than 20 percentage points in accuracy measured across complete groups of scenes.
AI Radar monitors Czech and international sources every day, looking for changes that truly deserve attention.
MonitorsOfficial AI company blogs, specialist media, and research sources.
Selects and combinesFilters out information noise and combines articles about the same change into a single event.
Summarizes and explainsExplains significant events in English: what happened, why it matters and where the information comes from.
The result is a quick overview of what has actually changed in the AI world, rather than another stream of articles.
Use the CS/EN switch to read the same Radar in Czech or English. English content is published after its translation has been checked, so new and older items may appear later.
Everything you need to navigate the AI world
Today’s briefingAn overview of news from the past 24 hours and a main story when available. Beside them is Live · AI Flash, followed by research and links to other Radar sections. On mobile, the blocks appear one below another.
AI FlashAn ongoing feed of brief updates with an evidence status. Links lead to a Radar detail page when one is ready, otherwise to the original source. You can also find reset and outage histories here.
Practical applicationsWhat new tools and features can do, what you can try and what their actual impact could be.
Model selectionModel comparison by type of work, capabilities, price and speed.
Research and archiveA separate research overview, topic search and older events by date.
One event, everything that matters
Each row represents one event — not one article. At a glance, you can see its significance, credibility and main point.
Illustrative example, not a current news item.
Importance: ▮▮▮ majorOpenAIModels✓ 6
Agent mode is available to all paying users
Until now, the mode was available only on the highest plan; it is now available on all paid tiers without a waitlist.
▮▮▮ major · ▮▮ important · ▮ we're tracking = how significant the change is✓ 6 = six independent publishers, not the number of articles or feeds✓ official = a clear release, law or incident is substantiated by the relevant authority1 source = no independent confirmation yetbold = who is behind the changegray text = a brief summary of what happened
The detail page contains a fuller summary, its significance and original sources. Practical impact appears in the detail and the For individuals and For businesses views. An AI Flash item reaches the main selection only after it has been expanded and meets the publication rules.
The same news, two practical uses
We first summarize each event in the same way for everyone. Based on those same facts, we then explain what the change means for your own use and what it could mean for how a company operates.
For individualsWhat you can use or try, how the change can help you at work and what to watch out for.
For businessesWhat impact the change could have on processes, costs, risks and other business decisions.
Today’s briefing is the same for everyone. Pages
For individuals and For businesses
can be found in the main navigation — they select only events relevant to the given use case.