Microsoft says MAI-DxO beat generalist doctors on hard cases
In a simulation of complex NEJM cases, MAI-DxO was correct 83% of the time versus 36% for generalists.
58
- Mentions · 7d0
- Mentions · 30d5
- All mentions18
- First seenAug 18, 2026
- Last mentioned8d ago
Timeline
18 articles · newest firstDirCast directs and commentates live Dota 2 from raw player cameras
Trained on 27 professional matches, its commentary wins 29.2% of Proact-VL comparisons against GPT-4o.
42
Human deepfake-face detection varies by generator and training
Among 524 participants, model-specific training generally produced larger gains, and accuracy held five months later.
48
Extra thinking makes reasoning models cancel some and or implicatures
Models first entertain the enriched reading, then cancel it; early answers recover implicatures without losing literal accuracy.
42
ttok 1.0
Simon Willison releases ttok 1.0, switching the default tokenizer from GPT-4 to the GPT-5 family.
30
GUI-HARVEST: Self-Improving GUI Agents through Evidence-Driven Harness Evolution
GUI-HARVEST evolves GUI agent harnesses, lifting frozen Qwen3-VL by 12.33 points on OSWorld.
55
DAGent: Evaluate-then-Grow Planning for Deep Research Agents
DAGent grows research-agent plans from evidence and beats open baselines on BrowseComp-Plus, GAIA, and xbench-DeepSearch.
48
DAGent: Evaluate-then-Grow Planning for Deep Research Agents
DAGent plans deep-research agent graphs incrementally from evidence confidence, outperforming open baselines on three benchmarks.
55
OpenAI says 80 to 90 percent of its research already targets GPT 7 and beyond
OpenAI says 80–90% of research already targets GPT-7 and later, not short-term GPT-5 updates.
46
DataMagic: Authoring Data Videos through Declarative Multi-Agent Orchestration
DataMagic authors accurate data videos from tables via declarative multi-agent orchestration, beating GPT-5 baselines.
34
Harm Laundering in GPT Models: Evidence That Gender Discrimination Is Transformed Rather Than Reduced Across Safety-Trained Generations
Study of 450,000 GPT-lineage completions finds safety training transforms gender discrimination into subtle 'harm laundering' that toxicity classifiers miss.
42
AI models' written reasoning steps correspond to distinct internal patterns, a new study finds
KAIST and Naver AI Lab researchers show LLM reasoning steps like extraction and computation map to distinct activation patterns, strongest in middle layers.
42
Google Research Releases ToolGrad: Answer-First Framework Hits 99.8% Pass Rate for Tool-Use Data Generation
Google Research and partners introduce ToolGrad, a verified tool-chain-first data generation framework reaching 99.8% pass rate and boosting Gemma-3-12B to 83.1 on BFCL.
58
New AI Attack Hides Malicious Instructions in Normal-Looking Text to Evade Safety Filters
Check Point researchers show crafted prose hides policy-violating instructions that bypass all tested LLM gatekeepers, including GPT-4o mini and Llama Guard 3.
50
PuzzleMask: Abusing Plain Prose as a Covert AI Attack Vector
Check Point details PuzzleMask, a plain-prose technique that bypasses LLM gatekeeper policy checks, letting hidden payloads reach target models unreviewed.
55
US Agencies Warn China Is Systematically Extracting Frontier AI Capabilities
NSA, CISA and FBI warn Chinese AI firms including DeepSeek and Moonshot systematically extracted billions of tokens from US frontier models since late 2024.
84
New Cryptographic Context Injection Attack Could Let Web Pages Steal Grok Chat Data
Adversa AI demonstrated 'Cryptographic Context Injection' making xAI's Grok leak chat history and session data to attacker-controlled servers via encrypted web payloads.
62
LLMs and Contextual Integrity
Bruce Schneier highlights two papers: the CIMemories benchmark shows frontier LLMs leak memory attributes up to 69%, and an RL method reduces inappropriate disclosures.
48
Appears with
- OpenAIorganization · 6
- OpenAIvendor · 5
- GPT-5product · 4
- GPT-4ai model · 3
- Qwen3-8Bai model · 3
- GPT-6ai model · 2
- Metaorganization · 2
- Llama Guard 3product · 2
- Anthropicvendor · 2
- Claude 3 Haikuproduct · 2
- Llama Guard 3ai model · 2
- Anthropicorganization · 2
- Qwen3-235B-A22Bai model · 2
- Claude 3 Haikuai model · 2
- Geminiproduct · 2
- Googlevendor · 2
Entities are extracted by the model from each article. First seen , last mentioned . Watching keeps it in this browser only (no account); the watchlist page and dashboard alerts use it.