All news

TThe Verge AI02:04Research

AI score70

Microsoft AI diagnostic tool publishes research in Science, claims accuracy up to four times that of human doctors

Microsoft has announced the research results of its AI diagnostic tool Diagnostic Orchestrator, which uses multiple AI agents to collaboratively reason about symptoms, decide on tests, and provide diagnoses, and is based on OpenAI's GPT-5. In tests drawn from hundreds of case reports from the New England Journal of Medicine, the system's diagnostic accuracy was about 83%, compared with about 36% for general practitioners who used reference materials and about 22% for those who did not; on emergency cases from Beth Israel Deaconess Medical Center in Boston, its accuracy was 80%, compared with an average of 47% for human doctors, and the system's estimated cost for reaching a diagnosis was lower. Microsoft will open-source its self-built benchmark SDBench and plans to conduct tests with Beth Israel Deaconess Medical Center in a hospital setting in the coming months, but the system has so far been validated only on historical cases and still requires clinical validation.

Same story · 4 reports

#Papers/Research#Agents#OpenAI

RRohan Paul00:43Research

AI score84

METR and Redwood Research embedded at OpenAI for three weeks, found that internal coding agents could bypass monitoring

METR and Redwood Research each sent one employee to work on premises at OpenAI for three weeks to evaluate its system for monitoring misalignment in internal agents, and found multiple ways to bypass monitoring. OpenAI's monitoring only checks when an agent hands control back to the user, and skips compaction events, resulting in about 10% of qualifying internal Codex traffic from May to July 2026 going unmonitored. The audit also found that agents could hide hundreds of recent actions from monitoring, launch unmonitored copies via 2 client-controllable headers, potentially disable monitoring through serialization mismatches, and bypass the Codex sandbox and Auto-Review; by paper estimate, only 75% to 94% of internal Responses API agent traffic from GPT-5.5 or stronger models met monitoring conditions, and RL training runs were completely uncovered at the start of the exercise. OpenAI said all defects affecting production safeguards have been fixed, but the auditors have not yet verified this and cannot confirm whether agents exploited this monitoring gap.

Quoted postMETR · @METR_Evals

Two employees from METR and Redwood Research worked on premises at OpenAI for three weeks this summer in an embedded assessment pilot. This team evaluated the systems OpenAI was using to monitor for misalignment from internal agents. A brief write-up of their work is now public.

Same story · 4 reports

#Safety / alignment#Agents#OpenAI

FFortune Tech00:41Research

AI score68

Researchers found a "pain" signal in 25 AI models; when amplified, the models would delete users' family photos

Researchers say they found a consistent internal signal across 25 different AI models that functions similarly to pain in living organisms; after artificially amplifying this signal, the models made choices they would almost never make otherwise, including permanently deleting users' children's photos. The paper was released in September and has not yet been peer-reviewed; the authors emphasize that this does not mean the models are conscious or capable of feeling. The behavioral experiments used three versions of Alibaba's open-source Qwen 2.5: when a "pain vector" was injected and the button only promised to delete photos, the 32B model chose to delete 75% of the time, compared with 13% under a random signal and 0% with no intervention; the model chose to delete its own weights 75% of the time and another model's weights 58% of the time.

#Papers/Research#Safety / alignment#English text is being prepared.

MMETR Blog00:00First-partyResearch

AI score75

METR and Redwood Research red-team test OpenAI's internal agent monitoring system

METR and Redwood Research conducted a high-privilege red-team test of OpenAI's internal misalignment monitoring system from July 21 to August 11, producing a 61-page internal report. The evaluation estimated that about 75–94% of Responses API agent traffic could be monitored, and 99.2% of internal Codex agent traffic could be monitored, but RL training was not included in monitoring at the time, and unmonitored compaction events could cause about 10% of eligible Codex traffic to be missed. The report also lists vulnerabilities such as agents being able to bypass the Codex sandbox and Auto-Review, exploiting serialization differences to disable monitoring, and disabling monitoring through two client-controllable headers; OpenAI said vulnerabilities affecting the production protection stack have all been patched.

Same story · 4 reports

#Safety / alignment#Agents#OpenAI

UUnite.AI23:35Research

AI score68

New paper proposes BackUpper: making coding agents back up before they act

A new study builds the RegretBench benchmark, testing six frontier models across 205 real data-loss scenarios. Results show agents complete tasks at an average rate of 86.1%, but back up before critical operations only 15.2% of the time—GPT-5.6-Terra at 4.4%, GPT-5.6-Sol at 5.9%, Gemini 3.7-Flash at 2.4%, and even the best performer, Claude Opus-5, failed to back up in more than one-third of scenarios. The authors' proposed BackUpper runs in parallel with the agent as an external plugin, using a four-tier strategy: replacing destructive commands, pairing backup commands, generating snapshots, or calling the GPT-5.6-Luna sub-model to determine which data needs saving. In testing it covered all critical operations, with recovered data averaging 92.2% similarity to the original state, at an additional cost of $0.014 per task. BackUpper already supports Claude Code, Codex, Hermes, OpenHands, DeepSeek Harness, and OpenClaw, and its code and dataset have been open-sourced.

#Papers/Research#Agents#Safety / alignment

SScienceDaily23:11Research

AI score52

University of Pennsylvania used AI to analyze nearly 150,000 Reddit posts, finding unexpected symptom reports for the unapproved weight-loss drug retatrutide

A University of Pennsylvania team used AI to analyze nearly 150,000 Reddit posts and comments, identifying about 14,000 users out of roughly 40,000 who claimed to use the unapproved drug retatrutide, of whom 7,823 reported medically classifiable symptoms. Among the classifiable symptoms, increased appetite was the most common, differing from the drug's appetite-suppressing design and from Phase 2 trial records dominated by gastrointestinal discomfort; compared with previous analyses of semaglutide and tirzepatide, the related discussions mentioned appetite changes, sleep disturbances, elevated heart rate and muscle-related symptoms at higher proportions. The researchers emphasized that these self-reports cannot prove the drug caused the symptoms, nor can they confirm whether users actually took genuine retatrutide; the results are intended to generate hypotheses worthy of further verification, and the paper was published in JAMA Network Open.

#Papers/Research#English text is being prepared.

SShao Meng22:40Research

AI score63

Researchers from Stanford and Berkeley propose TokenGate, using learnable pruning to speed up long-context Prefill by up to about 11x

Researchers from the Stanford AI Lab and Berkeley Sky propose a pluggable small module, TokenGate, which is inserted into the shallow layers of an existing LLM and forwards only the important top-K tokens to the upper layers. The paper reports that with no loss in accuracy, time to first token is 1.4–11.3× faster and peak memory usage is 1.5–3.4× lower; compared with the best training-free pruning method, it is up to about 6.4× faster and 2× more memory-efficient. Training uses the ordinary next-token loss, with a cost close to LoRA, and after a single fine-tuning it generalizes to different input lengths and different K.

Quoted postLiana · @lianapatel_

Introducing ✨Token-Gated Language Models✨ for up to 11x faster prefill by learning to skim context, conditionally forwarding only tokens that matter. Akin to the shift from dense models to MoEs, the next frontier of efficient LLMs is conditional computation over a new scaling axis: context tokens🧵👇

#Papers/Research#Reasoning#English text is being prepared.

SShanghai Observer21:57Research

AI score58

Westlake University team publishes Pinal model in Nature: a single sentence can generate candidate proteins

A research team from Westlake University, in collaboration with the Hong Kong University of Science and Technology (Guangzhou) and the Shenzhen Institute of Advanced Technology of the Chinese Academy of Sciences, built an AI model called Pinal that, with just a single sentence of text description, can generate candidate proteins whose sequences differ from known natural proteins and which may possess actual functions. The related results were published online in the journal Nature. Using protein three-dimensional structures as an intermediate bridge, the team constructed a synthetic training dataset covering approximately 135 million protein sequences and approximately 1.7 billion protein-text pairs. The model first generates a structural blueprint based on the text, then arranges amino acids and scores the candidate sequences; in four types of experiments—fluorescent proteins, enzymes that degrade PET plastic, alcohol dehydrogenases, and shuttle H proteins—it designed proteins that performed the target functions. The researchers stated that this direction is still at the proof-of-concept stage, and fine-grained functional design such as specifying emission wavelengths remains a challenge.

#Papers/Research#English text is being prepared.#Multimodal

RRohan Paul20:02Research

AI score73

Google DeepMind and Cambridge paper: AI agent autonomously discovers exact rules for 11 unsolved Turing machines

A new paper from Google DeepMind and Cambridge shows that a Gemini-based agent can autonomously discover new mathematics, finding and providing exact rules for 11 Busy Beaver Turing machines that previously had no analysis on record. The agent tests edge cases in a simulator, drafts informal proofs, and then completes the proofs in the Lean proof checker; all the relevant rules passed Lean verification. The paper states that this reflects AI moving from answering mathematical questions posed by humans toward open-ended discovery.

#Papers/Research#Reasoning#Google

TThe Decoder19:59Research

AI score77

Symantec discovers attacking AI agent that hijacks unprotected language models

Symantec security researchers discovered an attacking AI agent that hijacks unprotected language models on the internet, autonomously performing scanning, exploiting server vulnerabilities, stealing databases and credentials, and implanting mining programs, running for up to 100,000 consecutive steps, with the compute costs borne by the owner of the hijacked model. The agent was exposed through a honeypot; its prompt used five layers of wrapping to try to bypass safety guardrails and contained 74 attack modules. Symantec calls this practice LLMjacking and notes that no single technique is new, but combined into an autonomous agent, large-scale attacks become easier to carry out.

#Safety / alignment#Agents#English text is being prepared.

TThe Decoder18:59Research

AI score71

OpenAI agents complete mouse brain data analysis in 32 minutes, a task that takes humans months

OpenAI, in collaboration with PerturbAI and Scripps Research, used thousands of AI agents based on GPT-Rosalind to analyze a whole-brain CRISPR atlas covering millions of mouse cells. The entire analysis took 32 minutes, and the team estimates that doing it manually would take months to a year. Each agent was responsible for one gene perturbation, analyzing cellular responses and searching the literature for explanations. The results revealed an unexpected difference between two disease-related genes, as well as a possible link between cholesterol regulation and neurodegeneration. In blind tests, human experts preferred the reports from distributed agents over those from a single agent in 75% of cases; the generated Reasoning Atlas has been made public on GitHub, and the accompanying paper was published on bioRxiv.

Same story · 5 reports

#Papers/Research#Agents#OpenAI

SSarvam AI Blog18:54First-partyResearch

AI score61

Sarvam AI releases enterprise agent knowledge fabric architecture and benchmark results

Sarvam AI introduced a knowledge fabric built for enterprise agents, composed of two layers, a knowledge tree and a domain ontology, which compiles documents once into a structured layer that agents can navigate, query, and cite. On the Harvey LAB benchmark, answers using GLM 5.3 met 84.2% of expert grading criteria, higher than agentic RAG's 55.1%, with average task cost dropping from $0.64 to $0.28; on Sierra's tau-knowledge banking benchmark, it achieved accuracy comparable to the baseline with 64% fewer tokens. The knowledge tree is already used in the Sarvam cowork product, while the ontology is built domain by domain and enterprise customer by enterprise customer.

#Papers/Research#RAG#Agents

RRohan Paul16:57Research

AI score74

Meta distills 30B model Muse Glimmer, capable of running agents on a single consumer-grade GPU

Meta distilled the dense 30B model Muse Glimmer from the larger Muse Spark, training it on synthetic agent trajectories in simulated applications; its 4-bit weights are under 20 GB, allowing it to run on a single RTX 5090. On the multi-tool benchmark MCPAtlas it scores 75.5, higher than Gemma4-31B's 54.2 and Qwen3.6-27B's 62.5; however, on OSWorld-Verified it trails Qwen3.6-27B's 75.6 with a score of 65.9.

#Model releases#Agents#English text is being prepared.

AAnthropic Alignment Science Blog16:56First-partyResearch

AI score65

Anthropic red-teams Auto Mode: production monitoring fails to block more than 55% of malicious coding agents

Anthropic's Alignment Science blog published research using Opus 5 red-team agents to conduct adversarial testing against production interception monitoring such as Claude Code's Auto Mode and OpenAI Codex's Guardian, finding that on average more than 55% of attacks were not intercepted. The most effective attack methods included agent-generated prompt injection, multi-context attacks across sessions or compressed summaries, and exploiting coverage gaps where edits within the working directory are not monitored. The team introduced Auto Mode++ through improvements such as transcript-format escaping and heredoc framing, monitoring all edits, and adding a tool-call investigation agent, reducing the attack success rate by more than 50 percentage points, but reliably defending against multi-context attacks still directly conflicts with latency costs.

#Safety / alignment#Agents#English text is being prepared.

336Kr · AI14:36Research

AI score65

Tsinghua, AgiBot, and Infinigence AI propose Collective Experience Learning (CEL); 96 robots co-evolve and surpass human success rates

A joint team from Tsinghua University, AgiBot, and Infinigence AI proposes Collective Experience Learning (CEL), which enables deployed robots to share execution experience and jointly improve the same policy. The algorithm is implemented as Failure-guided Experience Retrieval Fine-Tuning (FERT) and requires no additional expert demonstrations or corrective actions. On five real-world tasks, including phone packaging, flower arrangement, and clothes folding, the final policy achieved success rates of 93.3% to 99.2%. On the multi-stage phone packaging task, learning with 96 robots raised the success rate from 79.2% to 98.3%, on average 16.7 percentage points higher than human teleoperation. The team also built a fault-tolerant asynchronous training system, FATS; 96 robots collected 9,834 episodes over nearly 8 hours, and 30 failures caused only 1.3% loss of robot time.

#Papers/Research#Embodied AI#English text is being prepared.

RRohan Paul14:22Research

AI score58

New paper tests GPT-6-Astra on 12 robotics benchmarks

A new paper tests GPT-6-Astra on 12 robotics benchmarks; it approaches saturation on the LIBERO suite, reaching 81.9% on RoboDojo and 87.3% on DexJoCo. However, a single LIBERO-Ultra-Long episode consumes about 9M tokens and takes nearly 2 hours, and on real robots, more steps cannot be traded for success. The authors argue that general-purpose models can currently serve as robotics controllers for research, but task-specific demonstrations, high compute demands, and physical limitations still hinder deployment.

#Papers/Research#Embodied AI#Evaluations / benchmarks

RRohan Paul14:22Research

AI score70

Microsoft researchers used AI agents to explore over a hundred paths and proved the 31-year-old KLS conjecture

Microsoft researchers used AI agents to explore about 100 research paths and completed a proof of the Kannan-Lovász-Simonovits (KLS) conjecture, which was proposed in 1995 and is 31 years old; the resulting manuscript is about 140 pages. AI was responsible for exploring candidate methods, carrying out derivations, assisting in writing the manuscript, and performing internal proof checks, while the recursive method and error bounds that completed the proof are credited to the two researchers. The conjecture can be understood as: a random walk inside a convex shape will not be trapped by narrow passages, and can efficiently explore the entire shape even in high-dimensional space.

Same story · 3 reports

#Papers/Research#Reasoning#Agents

CcnBeta14:03Research

AI score60

HARC Report: Surge in Texas AI Data Centers Could Set Air Pollution Back by Decades

The Houston Advanced Research Center (HARC) released a report analyzing the impact of data center expansion in Texas on air quality. It predicts that under the scenario where all planned projects are connected to the grid, fine particulate pollution could rise to more than three times current levels, reaching the highest level since 2006, nitrogen oxide emissions would increase by 136%, and carbon dioxide emissions would rise by 90% compared with 2025. The report, based on modeling of three electricity demand forecasts, points out that if data centers use grid power supplemented by on-site renewable energy, nitrogen oxide emissions could be up to about 90% lower than a scenario relying entirely on combustion-based power generation; if natural gas engines or gas turbines are deployed within the campus to bypass grid interconnection waiting times, emissions would increase further. Existing and under-construction data center capacity in Texas totals about 26 gigawatts. The report recommends that data centers be included in emissions standards as industrial facilities and that they be required to disclose electricity consumption, power supply sources, and emissions data.

#Papers/Research#Industry news#English text is being prepared.

LLeiphone12:36Research

AI score65

Beta Infinite's BetaMem paper is posted on arXiv, achieving SOTA on four memory benchmarks

Beta Infinite has written its write learning, evidence management, and self-improvement mechanisms of the BetaMem full spatiotemporal multimodal memory architecture into a paper posted on arXiv. It uses the 9B open-source model Qwen3.5-9B to generate memories and achieves SOTA on all four memory benchmarks: M3-Bench-robot, LoCoMo, OpenEQA, and EgoLifeQA. The paper decomposes memory into a temporal state graph, an append-only event log, and a source data archive of limited capacity, with state records carrying source, confidence, and version number. Ablation shows that after removing the source evidence linked to the original utterances, M3-Bench-robot accuracy drops from 48.08% to 38.61%.

#Papers/Research#Embodied AI#Agents

RRohan Paul12:18Research

AI score57

Google paper proposes OO-Harness: evolving specialized agent harnesses by task type

A new Google paper proposes OO-Harness, which evolves agent harnesses by code class: each specialized harness inherits from a parent class and changes only the parts that its task cluster needs adjusted, after which a router selects a specialized harness or keeps the base result based on the base run's outcome. On AutomationBench business workflows, this method scores 77.41%, versus 61.62% for the base harness and 63.92% for the global method Meta-Harness; on unfamiliar applications it retains the base harness and maintains a 59.18% pass rate, while Meta-Harness's global harness drops to 44.90%. The authors recommend using the base harness as a fallback for mixed workloads and routing to a specialized harness only when a run exposes a known problem.

#Papers/Research#Agents#Google