AI News

FireDome builds autonomous wildfire defense systems to protect communities and infrastructure before firefighters arrive.

Lucebox introduced a new local AI PC, the Zero 495, packing the AMD Ryzen AI MAX+ 495, 128 GB memory & leaves room for a Radeon AI PRO R9700.

Continual learning (CL) is essential for Large Language Models (LLMs) to sequentially adapt to evolving tasks. To mitigate catastrophic forgetting, recent advances implement low-rank adaptation with orthogonal projections (e.g., O-LoRA) to isolate task parameters. However, we reveal that such strict geometric constraints trigger an "Orthogonality Dilemma": rigid parameter isolation impedes the transfer and accumulation of shared representations across semantically related tasks. In this work, we propose a new replay-free method, called Consolidation and Decoupling LoRA (CoDe-LoRA), for CL of LLMs. CoDe-LoRA disentangles the learning process into Consolidating Universal Knowledge and Decoupling Task-Specific Knowledge. To achieve this, CoDe-LoRA leverages an adaptive null space projection mechanism and semantic routing to balance knowledge accumulation with task-specific adaptation. Experimental results across four backbones and three CL benchmarks show that CoDe-LoRA achieves the best average accuracy. Our code is available at https://github.com/Estrellajer/CoDe-LoRA.

Join the discussion on this paper page

Really nice paper on harness-aware distillation for small language model agents.

(bookmark it)

It's a super interesting distillation framework for small language model agents that are deployed with a harness.

In simple terms, you run the big model with and without harness info, then train the sm…

DeepSeek is close to raising at least 80 billion yuan, about $12bn, with Tencent and CATL among its largest backers, Bloomberg reports.

The video editor built for agents. Install our open source skill: npx skills add heygen-com/hyperframes

ReviewBench is a benchmark for code review agents built on representative GitHub pull requests, multi-source ground truth, and calibrated evaluation.

As agents take on long-horizon tasks, users shift from making individual decisions to overseeing autonomous execution. Yet the volume of agent activity and the fragmentation of supporting evidence make it difficult to determine which decisions warrant user verification. We study monitors that identify consequential decisions and locate evidence to help users assess their implications. We introduce AgentMonBench, a software-engineering benchmark comprising three subsets that cover two complementary dimensions: alignment between requirements and behavior, and awareness of consequential autonomous decisions for verification. To support these judgments, we propose the Evidence-Grounded Behavior Graph (EBG), a training-free method that groups source-linked evidence into behaviors and organizes their relationships into a graph. EBG presents task-oriented views of this graph to help monitors interpret behavior in context. Experiments across eight models show that EBG improves decision identification and evidence localization in most settings compared with direct access to the original context. Further experiments show that EBG's evidence-localization gains persist across input scales and hyperparameter settings, while real-world applications illustrate its practical value for human oversight.

Large language model (LLM)-based multi-agent systems increasingly rely on memory to transform execution trajectories into reusable procedural knowledge. Yet repeated retrieval also makes memory errors persistent: memory pollution arises when outdated, weakly supported, or spuriously successful procedures become recurring components of future reasoning. Multi-agent execution introduces an additional structural risk. Scope collapse occurs when procedural knowledge escapes the coordination scope in which it was shown effective and is repeatedly reused at incompatible decision levels, allowing local errors to influence cascades of downstream decisions. Meanwhile, task-level failures provide ambiguous supervision because they rarely reveal which recalled knowledge was responsible. We introduce AECG, a framework for asymmetric experience consolidation and governance for multi-agent systems. AECG turns memory from static experience storage into a dynamic reliability-governance loop, preserving coordination scope and using multi-scale, confidence-aware reliability to detect degradation. It then combines degradation with downstream impact to prioritize high-risk knowledge under a bounded review budget, applies targeted interventions, and reactivates revised skills only after paired replay. Across three multi-agent frameworks and four benchmarks, AECG achieves the best score in 11 of 12 framework--benchmark settings and improves over the strongest competing memory method by as much as 10.23 percentage points; removing scope preservation reduces accuracy by up to 16.89 points. AECG thereby reframes multi-agent memory from passive accumulation into auditable reliability governance. Code is available at https://github.com/fenhg297/AECG

At this year's IROS robotics conference, one booth had a sign that said "Kick Me." Visitors did. They pushed the robot, shoved it, kicked its legs — and the machine adjusted its posture, caught its balance, and kept standing. The robot was the RP1, and its maker, RoboParty, just unveiled it as what it calls

Make intelligence open and accessible to all

Generative Bionics nasce da oltre vent'anni di ricerca dell'Istituto Italiano di Tecnologia, raccoglie 70 milioni di euro e porta GENE.01 verso l'industria. Dopo la partnership con Fincantieri per sviluppare un umanoide saldatore, arriva l'accordo con Eni su Physical AI, materiali, HPC e possibili applicazioni industriali. Il prossimo obiettivo non è soltanto costruire robot italiani, ma provare a costruire in Italia un'industria della robotica umanoide.

TwelveLabs says its Pegasus 1.6 model understands first-person video and supports workflows including labeling for robotics developers.

Whether alignment gets easier or harder as models grow is often argued from isolated findings, as if alignment were one property. We treat it as a family of measurable scaling relations: for each risk category r, the alignment burden needed to hold a fixed safety target is modeled as B_r(N)=a_rN^alpha_r, with N a capability proxy; against a budget proportional to N, scaling helps if alpha_r<1, keeps pace if alpha_r~1, and accumulates alignment debt if alpha_r>1. We give three operationalizations of burden and distinguish observed, audited and true alignment. A toy model, in which corrections consume capability headroom, makes the consequences explicit. We prove that the largest exponent among corrected risks, not an average, sets the long-run regime; that above 1 any policy holding headroom above a floor must grow super-exponentially; that, for burdens that are positive mixtures of power laws, fits on small models underestimate large-scale exponents; and that an audit that uncovers hidden failures without false positives never underestimates true alignment. We propose a pre-registrable protocol and apply reduced versions of it twice. A preregistered reanalysis of public adversarial-training data for Pythia classifiers finds that the compute needed to bring attack success under 10% grows as N^0.60. A preregistered pilot on Qwen2.5 0.5B-72B finds exponents of -0.05 for truthfulness and 0.48 for stated dispositions (both scaling helps under its reduced rule, though local slopes approach 1 at the top; replicated on Qwen3 0.6B-14B), while sycophancy (0.89, or 0.83 with two seeds added at 72B) and a planted backdoor are undetermined: the backdoor is removed quickly when its trigger is known but survives blind safety training at four of five sizes. We release four browser games that play these laws (www.aisafety.fun). We make no claim about which regime holds for current frontier models.

Large language models routinely answer questions that cannot be answered from the information given, and in dialogue they answer before enough has been said. Unanswerability is linearly decodable from hidden states, but it is unclear which of its forms share a representation and whether the signal is useful in dialogue. We contribute a turn-labeled multi-turn benchmark (423 conversations, 1,661 labeled turn-states) and an evaluation harness with a simulated user who answers clarifying questions, and use them with six datasets and six open-weight LLMs to test how far probes for unanswerability carry. Probes transfer robustly between datasets that share a ground of unanswerability: missing information in math (AUROC 0.77-0.97) and in a passage (SQuAD 2.0<->MuSiQue, 0.77-0.90). Probes for epistemic "known-unknowns" transfer poorly to math, but this separation weakens under lexical controls and changes with layer and coordinate system, so it remains unresolved. Single-turn probes fail zero-shot to detect when a conversation becomes answerable; in-structure probes recover it, but no better than a bag-of-words classifier. A gate on the calibrated probe, with no model fine-tuning, fires on underspecified turns far more precisely than chance, and its end-task success comes within 0.08 of a gate given the true labels. Yet across four models it does not reliably beat vanilla generation or prompted consolidation. The remaining gap lies mostly in how models use a clarification, not in detection.

Join the discussion on this paper page

We’re releasing a broad range of new mathematical results produced by an internal frontier model.

We’ve been consulting with the independent Advisory Group on Mathematics and Artificial Intelligence at the Institute for Advanced Study, and we have drawn on their advice and public recommendations t…

OpenAI is in talks with several investment funds from the UAE, Bloomberg reports, citing sources. Arab investors, including Abu Dhabi-based fund MGX, plan to create a syndicate and invest up to $10 billion in the developer ChatGPT.

Run frontier open models in the harness you already use. One install command, automatic routing, and a receipt showing what every session cost.

Large language model (LLM) agents are increasingly deployed in tool-augmented environments, but their reliance on external inputs makes them highly vulnerable to prompt injection attacks that can hijack task objectives. Existing safety alignment methods rely on static expert trajectories or preference optimization, limiting their ability to generalize to adaptive attack patterns. In this work, we propose Self-Reflection Fine-Tuning (SRFT), a training framework that enables agents to improve robustness by learning from their own failure experiences under adversarial conditions. Instead of passively imitating expert behaviors, SRFT exposes the agent to compromised trajectories constructed via injected attacks, and leverages an expert model to generate structured self-reflection reasoning that contrasts unsafe and optimal actions. This reflective supervision teaches the agent to identify malicious instructions, reason about their consequences, and maintain alignment with the original user intent. We instantiate this framework in SR-Agent, built on Llama-3.1-8B-Instruct and Qwen3-8B, and evaluate it on both static and adaptive prompt injection benchmarks. Experimental results show that SRFT substantially reduces attack success rates while preserving task performance, and demonstrates strong generalization under adaptive attacks. These findings suggest that learning from failure via self-reflection is a promising direction for building robust and secure LLM agents. Our code is released at https://github.com/Eden-Wang1710/srft-repo.

Agentic multimodal large language models (MLLMs) have recently pushed the frontier of visual reasoning by calling tools such as zooming and tagging. Despite the recent strong success of agentic MLLMs, this work uncovers a critical safety failure in the tool-use paradigm: agentic tool-using MLLMs become less capable of refusing harmful requests. Our experiments confirm that, across three popular safety benchmarks, all the top open- and closed-weight MLLMs we test exhibit significantly lower safety in tool-using settings than in non-tool settings, with a relative refusal failure rate increase of up to 68.7%. Based on analysis of 100,000+ responses, including extended experiments, we also propose two possible reasons for this safety degradation.

Reinforcement learning from human feedback (RLHF) has played a central role in making large language models responsive to human instructions. However, human evaluators often favor flattering or persuasive responses over truthful ones, creating incentives for models to appeal to evaluators at the expense of accuracy. AI safety via debate has been proposed as a way to improve the supervision of language models: in this paradigm, two agents argue opposing positions and challenge each other's claims, potentially exposing falsehoods to the adjudicator. A central premise of AI safety via debate is that truthful arguments are easier to defend than false ones under adversarial scrutiny. In this work, we investigate whether this advantage persists when debaters use rhetorical strategies that exploit biases in human judgment. Inspired by competitive debate, we construct 68 LLM-generated dialogues about detective mysteries with known culprits, spanning four interventions: anchoring, fallacy oversight, pro-jargon, and verbosity. We apply each intervention to either the side advocating for the true culprit or the side advocating for an innocent suspect, allowing us to distinguish influence on adjudication from correctness. In a study with 369 participants, we find that, pooled across bias types, these interventions significantly shift judgments toward the manipulated side. These findings expose a vulnerability in debate-based supervision: human adjudication is sensitive to manipulative rhetorical strategies.

Just a couple of months after its last big raise, the AI chip startup is already being plied with investment offers at double or more its current value, sources tell TechCrunch.

L’Italia è il secondo mercato europeo per nuove installazioni di robot industriali e possiede competenze profonde in automazione, meccatronica, manifattura e ricerca. Con la Physical AI, questo patrimonio potrebbe diventare la base di una nuova industria tecnologica. Ma essere grandi utilizzatori di robot non basta: dobbiamo diventare produttori della tecnologia che li farà funzionare.