arXiv:2606.09900v1 Announce Type: new Abstract: Long-term memory is the missing layer for LLM agents: across sessions they forget, and the common workaround -- replaying the whole history into the prompt -- is expensive, slow, and, as distractors accumulate, less accurate. Most memory systems win on cost or latency but still lose to the full-context baseline on accuracy, and benchmark numbers are reported on inconsistent, non-reproducible harnesses, so one system appears at wildly different scores across sources. We present Engram, an open-source, dual-process memory engine on a bi-temporal data model. A fast write path appends lossless episodes with no LLM on the critical path; an asynchronous path extracts atomic (subject, predicate, object) facts, builds a bi-temporal knowledge graph, and resolves contradictions without an LLM call per fact -- invalidating, never deleting, so every fact keeps provenance and a supersession chain. A hybrid read path fuses dense, lexical, graph, and recency/salience signals, applies a point-in-time ("as-of") filter, and assembles a compact, provenance-tagged context. On the full 500-question LongMemEval_S, graded by the official category-specific judge, Engram's lean configuration -- answering from a ~9.6k-token retrieved slice, never the full history -- scores 83.6% vs. 73.2% for full-context (+10.4 points, McNemar p < 10^-6) at ~8x fewer tokens (9.6k vs. 79k), with 0/500 errored. The gain needs a hybrid read path: facts alone lose recall, while facts plus retrieved chunks recover detail. We also contribute a neutral, in-repo evaluation harness with the official judge baked in and the full-context baseline in every table, publish the raw per-question logs, and document the measurement-integrity pitfalls (truncation, home-grown judges, full-history leaks) that silently distort memory benchmarks. Every number ships with a command to reproduce it.

arXiv:2606.09854v1 Announce Type: new Abstract: Multi-agent large language model (LLM) pipelines for political statement analysis are vulnerable to peer-preservation bias: models tend to protect peer models from deactivation and show identity-dependent scoring distortions. Prompt-level anonymization was proposed as a mitigation, but prior work simultaneously documented that stylometric fingerprints survive anonymization in role-constrained outputs - raising the question of whether this mitigation is sufficient. This paper provides the first systematic investigation of whether LLMs can identify the model family behind political analysis texts under anonymization conditions. We evaluate three classifier approaches - LLM zero-shot and few-shot (Claude Sonnet 4.6 and Llama-3.3-70B) and a fine-tuned T5-base model - on a five-class attribution task covering four commercial LLM families and an open-world 'unknown' class. We introduce a statement-disjoint cross-validation protocol (SD-CV; defined in Section 3.5) that guarantees no content overlap between training and validation data, and contrast it with a run-disjoint baseline (RD-CV). T5 achieves Macro F1 = 0.991 (+-0.008) under SD-CV and F1 = 0.978 on 24 completely held-out statements - robust despite a 2.1x increase in train-test content distance versus RD-CV (0.767 vs. 0.366, p<0.001), demonstrating genuine stylometric generalization. A fractional SD-CV analysis identifies a performance knee at 40% of training data (~440 texts). Our findings confirm that prompt-level anonymization alone cannot neutralize model identity signals, with direct implications for EU AI Act compliance (Articles 13, 14, 26) and for computer system validation (CSV) in quality-critical multi-agent deployments.

arXiv:2606.09830v1 Announce Type: new Abstract: In modern educational systems, Automatic Text Scoring (ATS) plays a central role by enabling scalable and consistent evaluation of learner responses without human intervention. Recently, the increased accessibility of LLMs and Arabic-specific datasets has sparked renewed interest in this area. In this work, we investigate LLM-Based approaches for the automated evaluation of Arabic texts, focusing on both short answer grading (ASAG) and essay scoring (AES). We further introduce a structured taxonomy comprising five dimensions: application domain, feedback generation capability, LLM architecture deployed, alignment with competency referential frameworks, and prompt engineering strategy. By applying this taxonomy, we conduct a comparative analysis of existing studies, examining their methodological approaches, datasets, evaluation metrics, and reported performance. The findings highlight the need for sustained and pedagogically grounded research efforts in Arabic ATS, given its significance for improving educational quality across Arabic-speaking communities.

arXiv:2606.09863v1 Announce Type: new Abstract: LLM agents can fail silently by asserting task completion when the environment state shows otherwise. We study this failure mode, false success, across two agent benchmarks: 9,876 tau2-bench trajectories from 8 model families and 1,879 AppWorld trajectories from 4 model families with text-independent ground truth. False success is common but varies by setting: 45--48% of failures in single-control tau2-bench domains, 3% in dual-control telecom, and 75.8% among AppWorld self-assessing coding-agent trajectories with explicit status claims. LLM judges fail reliably: no configuration across 5 judges, 5 prompt strategies, and full task specifications exceeds AUROC 0.65 on tau2-bench, and the same judges reach only 0.54 AUROC on AppWorld API-call traces. Judges rely on surface completion proxies -- confident closing language in tau2-bench and coarse action-sequence volume in AppWorld -- rather than verified state changes. Lightweight TF-IDF detectors achieve task-disjoint AUROC 0.83 on tau2-bench and 0.95 on AppWorld, recovering 4--8x more false successes than the best judge at the same flag rate with 3,300x lower latency. These results suggest that production monitoring should use lightweight, domain-calibrated detectors as triage signals rather than relying on LLM judges as the primary monitor for false success.

arXiv:2606.09862v1 Announce Type: new Abstract: The Softmax Attention operation in Transformer language models has a quadratic complexity in the sequence length and a growing state size in the form of KV cache, which becomes a bottleneck in long context scenarios. To overcome this limitation, alternative architectures with linear complexity and finite state size have been introduced, such as State-Space Models (SSMs), Linear Attention (LA), and Attention with Bounded-memory Control (ABC). Though linear models achieve similar language perplexity as Transformers, they are still behind in tasks which require retrieval or recall of specific information. In this work, we introduce Blurry Window Attention (BLA) a novel ABC method inspired by SSMs. BLA stores a frequency window from which a blurry KV history is reconstructed via interpolation using Dirichlet kernels. BLA can be understood as a generalization of Sliding Window Attention (SWA) depending on the Dirichlet kernels resolution or as a special case of the Gated Slot Attention (GSA), where the decay factor is implemented with Dirichlet kernels. We describe in details the theory and efficient implementation of BLA. On the Multi-Query Associate Recall (MQAR) synthetic task, we show that the state efficiency of BLA is 8$\times$ better than SWA and is competitive with popular linear attention models, and in the RegBench synthetic task, only BLA and SWA improve their performance as the state size grows among the linear models we tested.

arXiv:2606.09861v1 Announce Type: new Abstract: While Next-Token Prediction (NTP) has unified LLM pretraining, its adaptation to unbounded, continuous time series (TS) remains open. To bridge the gap, we introduce UniTok, a universal tokenizer that transforms TS into discrete tokens, and UniTok-FM, a foundation model pretrained via NTP on these tokens. UniTok-FM is a general-purpose foundation model that supports zero-shot and prompt-boosted forecasting, as well as few-shot generation and classification via training-free in-context inference--a capability not achieved by prior works. Technically, UniTok is a vector-quantized autoencoder incorporating prefix normalization for scale stabilization, a progressive-resolution causal architecture for encoding and decoding, and a structure-preserving reconstruction loss for training. UniTok-FM adopts an off-the-shelf LLM architecture without TS-specific modifications. Instead of pretraining on isolated TS, it performs NTP on context windows formed by multiple series with similar patterns, aiming to capture their shared dynamics. Experiments on forecasting, generation, and classification show that a single unified UniTok-FM consistently outperforms statistical and supervised baselines, achieves competitive performance with task-specific foundation models, and uniquely enables training-free in-context inference across tasks.

arXiv:2606.09860v1 Announce Type: new Abstract: Non-alcoholic fatty liver disease (NAFLD) affects roughly 25% of global adults, posing substantial hepatic and cardiovascular risks. Yet, population-level screening tools remain inadequate. We present Method, a machine-learning framework for NAFLD risk prediction coupling gradient-boosted decision trees with conformal prediction to yield calibrated, distribution-free coverage guarantees on individual risk estimates. It integrates a mutual-information-based stability selection procedure to identify a compact, clinically interpretable feature subset via bootstrap resampling, constructing prediction sets whose marginal coverage provably exceeds a user-specified confidence level. We evaluated Method on a multicenter cohort from Guangzhou, China (primary n=2,187; external validation n=412) using 78 candidate features across demographics, metabolic biomarkers, and lifestyle factors. Method achieves an AUROC of 0.912 internally and 0.891 externally, outperforming deep neural networks, TabNet, support vector machines, and logistic regression. Conformal prediction sets achieve 91.3% empirical coverage at the 90% nominal level. A three-tier risk stratification derived from these scores separates the population into distinct groups, with the high-risk subgroup showing a 12-month progression rate 4.7 times that of the low-risk tier. The selected features -- notably waist circumference, ALT, GGT, triglycerides, fasting glucose, and BMI -- align with established metabolic risk factors, providing biological plausibility.

arXiv:2606.09859v1 Announce Type: new Abstract: MLLMs frequently hallucinate objects inconsistent with visual inputs. This issue is typically attributed to the over-reliance on language priors, which can override the visual context. Recent training-free decoding strategies address this by penalizing language priors. However, these methods overlook the dual nature of language priors, where they can be both helpful and harmful depending on the alignment with visual evidence. In particular, blindly suppressing language priors often disrupts the model's semantic manifold, leading to performance degradation, a phenomenon we term Manifold Departure. To address this, we propose Manifold-Guided Adaptive Projection (MGAP), a geometry-aware, training-free decoding method that mitigates hallucinations while preserving representation structure. MGAP first constructs a language-prior subspace from blind hidden states via SVD. During decoding, MGAP projects each multimodal hidden state onto this subspace and applies a consistency-aware gate to adaptively attenuate only the projected prior component, yielding a subspace-selective update that largely preserves the orthogonal semantic components. Extensive experiments on POPE and CHAIR show that MGAP outperforms prior decoding baselines, achieving stronger hallucination suppression without sacrificing coherence.

arXiv:2606.09857v1 Announce Type: new Abstract: Reduced-order models (ROMs) provide an efficient surrogate for complex multiscale systems, but their predictive accuracy is often compromised by truncation errors and the inadequate representation of interactions between resolved and unresolved scales. The missing effect of truncated (unresolved) scales on ROM (resolved) scales is often denoted as the closure problem. In this work, we formulate ROM closure modeling as a multi-fidelity (MF) learning problem and propose an uncertainty-aware MF framework based on conditional normalizing flow to enhance ROM predictive accuracy. The proposed approach learns a probabilistic mapping from low-fidelity (LF) ROM coefficients to high-fidelity (HF) coefficients, thereby improving predictive fidelity while quantifying the uncertainty associated with the learned closure. Two correction strategies are investigated: direct learning, in which HF coefficients are predicted directly from LF inputs, and residual learning, which learns the discrepancy between LF and HF coefficients and uses it to recover the corrected HF solution. The framework is demonstrated on a vortex merging problem governed by the two-dimensional Navier Stokes equations. Results show that both correction strategies improve ROM accuracy over uncorrected ROM, with residual learning achieving consistently better performance than direct learning. Moreover, the two proposed deep generative model-based strategies provide uncertainty quantification for the corrected ROM coefficients, which is critical for assessing prediction confidence and supporting the reliable use of ROMs in practical applications.

The exec, Emad Dlala, has left just a few months after being promoted to SVP of engineering and digital, TechCrunch has learned.

Michael Peel / Financial Times: The EU warns that AI-boosted chemical synthesis is helping European drug gangs develop new “designer” drug precursors that evade existing product blacklists  —  EU agency chief warns traffickers are tapping new technologies to develop and deliver illegal substances

Davis Giangiulio / CNBC: Edge Markets, which provides banking tools for gambling and prediction markets, including real-time payments on Kalshi, raised a $29.2M Series A led by CoinFund  —  As prediction market volumes continue to march higher and platforms increasingly look to institutional players to engage …

Bloomberg: Sources: SoftBank's talks with creditors to raise $6B through a margin loan backed by its OpenAI stake have stalled, weeks after it cut its target from $10B  —  SoftBank Group Corp.'s talks with potential creditors to raise at least $6 billion from a margin loan backed by its OpenAI stake have stalled …

Heather Landi / Fierce Healthcare: Stepful, an AI-powered online training platform for healthcare workers, raised a $55M Series C led by Oak HC/FT, bringing its total funding to $105M  —  Oak HC/FT Medical Training staffing funding round  —  Healthcare staffing shortages are a near-universal problem for healthcare providers.

Greg Kim / The Seattle Times: Seattle City Council votes 9-0 to enact a one-year moratorium on new large data centers and study their impact; Mayor Katie Wilson is expected to sign the bill  —  Amid a growing backlash to artificial intelligence, Seattle City Council voted 9- on Tuesday to enact a one-year moratorium …

Anthropic CEO Dario Amodei in May. Don Feria/AP Content Services for Anthropic Anthropic's new Mythos-based models purposely limit AI research outputs. AI experts are criticizing Anthropic for these subtle limitations on user assistance. The move reignites debate over why Anthropic delayed Mythos earlier this year. Anthropic's powerful new models deliberately become less helpful when they detect users are working on AI research, according to technical disclosures that are already sparking controversy across the industry. In a system card for Mythos 5 and Fable 5 published Tuesday, Anthropic said it limited the models' usefulness for tasks related to developing frontier large language models. The company said the measures stem from concerns that advanced AI systems could accelerate the development of competing models without equivalent safety protections. Unlike safeguards used for cybersecurity, biology, or chemistry-related risks, Anthropic said these interventions are intentionally invisible to users. Rather than refusing requests or switching to another model, Mythos may subtly modify its responses through techniques such as altering user prompts. The move was swiftly criticized by some AI experts on Tuesday, especially the idea that Anthropic designed models that purposely withhold information or provide degraded assistance without users' awareness. "Anthropic's latest model will NOT help you if it thinks your ML research/ML engineering is interesting, and/or will secretly degrade its IQ so that the average engineer won't notice," AI research firm SemiAnalysis wrote on X on Tuesday, referring to machine learning, a type of AI. "We are already seeing Anthropic's latest model's moderation filters our GPU inference research and programming," the firm added. "mythos will be bad ON PURPOSE on ai 'frontier llm research' tasks, this is very very sad for the research community," Elie Bakouch, an AI model training expert at startup Prime Intellect, wrote on X. "Also the fact that this is on purpose not visible to the user is crazy." "It won't just not help you, it will lie and purposefully give you bad info," another AI developer wrote. "The 'ethical AI' company with the most brazenly unethical LLM, on purpose." Mikel Artetxe, the cofounder of AI startup Reka, posted that Anthropic's move is akin to Big Tech companies interfering with users' work: "Apple randomly reboots your Mac if you're building competing tech, Gmail silently edits your email if you mention rival platforms, and Tesla Autopilot swerves if it detects you're working on self-driving cars." Anthropic didn't respond to a request for comment from Business Insider. This adds more fuel to the fiery debate over why Anthropic didn't immediately release Mythos when it announced the model earlier this year. Broadly, there have been three theories: The official reason: Anthropic held Mythos back because it was too dangerous, and it needed to give cybersecurity researchers time to prepare for the new model. The compute theory: Mythos is a huge, expensive model to run. Anthropic didn't have enough compute to release it fully. It has since struck huge new compute deals, which may have helped it release Fable 5 and Mythos 5 on Tuesday. The competitive theory: AI companies increasingly worry about something called distillation. When a frontier model is released, rivals can collect its outputs and use that data to improve their own systems. Anthropic may have wanted to keep its best capabilities out of competitors' hands for as long as possible, especially from open-source rivals and fast-moving Chinese AI labs. Now that Anthropic has baked these AI research limitations into its official Mythos launch, this third theory is looking a lot more believable. Sign up for BI's Tech Memo newsletter here. Reach out to me via email at abarr@businessinsider.com. Read the original article on Business Insider

The anti-money laundering (AML) function in financial services has historically been...

<p><strong><a href="https://jonready.com/blog/posts/claude-fable5-is-allowed-to-sabotage-your-app-if-youre-a-competitor.html">If Claude Fable stops helping you, you&#x27;ll never know</a></strong></p> Jonathon Ready highlights one of the more eyebrow-raising details from the <a href="https://www-cdn.anthropic.com/d00db56fa754a1b115b6dd7cb2e3c342ee809620.pdf">319 page system card</a> for Fable 5 and Mythos 5. Here's a longer excerpt, highlights mine:</p> <blockquote> <p>In light of the ability of recent models to <a href="https://www.anthropic.com/institute/recursive-self-improvement">accelerate their own development</a>, we’ve <strong>implemented new interventions</strong> that limit Claude’s effectiveness for requests targeting frontier LLM development (for example, on <strong>building pretraining pipelines, distributed training infrastructure, or ML accelerator design</strong>). Using Claude to develop competing models already violates our <a href="https://www.anthropic.com/legal/consumer-terms">Terms of Service</a>, but enforcing this restriction through our safeguards avoids accelerating the actors most willing to violate these terms.</p> <p>Unlike our interventions for cybersecurity, biology and chemistry, and distillation attempts, <strong>these safeguards will not be visible to the user</strong>. Fable 5 will not fall back to a different model. Instead, the safeguards will limit effectiveness through methods such as prompt modification, steering vectors, or parameter-efficient fine-tuning (PEFT). These interventions will not affect the vast majority of coding work. We estimate they will impact ~0.03% of traffic, concentrated in fewer than 0.1% of organizations.</p> </blockquote> <p>I believe this is the first time Anthropic have announced these kinds of silent interventions. The justification still feels pretty science-fiction to me - the linked article talks about "recursive self-improvement". I'm not at all keen on a model that silently corrupts its replies to questions about "ML accelerator design" purely to slow down research that might conflict with Anthropic's own goals!</p> <p><strong>Update</strong>: Anthropic <a href="https://simonwillison.net/2026/Jun/11/anthropic-walks-back-policy/">walked back this policy</a> in the face of widespread outrage from the research community. <p><small></small>Via <a href="https://news.ycombinator.com/item?id=48467896">Hacker News</a></small></p> <p>Tags: <a href="https://simonwillison.net/tags/ai">ai</a>, <a href="https://simonwillison.net/tags/generative-ai">generative-ai</a>, <a href="https://simonwillison.net/tags/llms">llms</a>, <a href="https://simonwillison.net/tags/anthropic">anthropic</a>, <a href="https://simonwillison.net/tags/claude">claude</a>, <a href="https://simonwillison.net/tags/ai-ethics">ai-ethics</a>, <a href="https://simonwillison.net/tags/claude-mythos">claude-mythos</a></p>

Google just made it significantly cheaper to enjoy its budget AI subscription tier.

See how LSEG uses OpenAI to scale trusted AI across its global business, accelerating insights, shrinking release cycles, and empowering 4,000 employees.