Your Edge Provider Now Handles Your Money: Inside Cloudflare’s Payment Consolidation Your website generates traffic from … Cloudflare’s Payment Layer: The Quiet Shift Toward Infrastructure-Native Monetization Read more The post Cloudflare’s Payment Layer: The Quiet Shift Toward Infrastructure-Native Monetization  appeared first on .

Discussion | Link

Cristina Criddle / Financial Times: Sources: Google DeepMind, Anthropic, and Meta have recently hired experts in psychology, ethics, and philosophy as they expand machine consciousness research  —  Google DeepMind, Anthropic and Meta study whether AI could become conscious — and what follows for humans if it does.

The US state of Florida has filed a lawsuit against OpenAI and Sam Altman, alleging the tech giant and its CEO put profit over public safety with its flagship artificial intelligence (AI) product, ChatGPT. The lawsuit, filed in Florida state court on Monday local time by Florida’s attorney general James Uthmeier, is one of the most significant enforcement actions brought by a state attorney against an AI company to date. It comes as OpenAI and other big tech companies are embroiled in a growing number of legal cases related to the alleged harm their products have caused. Six key elements The complaint opens with a screenshot of OpenAI’s own parental-control page, which states that ChatGPT was “built with safety in mind”. Then, in a standalone paragraph, the State answers with two words: “Not so”. This signals the central allegation of the case: that OpenAI sold ChatGPT to the public as safe and reliable, while knowing it could cause serious harm More specifically, there are six key themes to Florida’s case against OpenAI. The first is that the company engaged in deceptive safety marketing, assuring parents the platform is safe for teenage use, while not clearly disclosing that ChatGPT can be wrong. Second, despite OpenAI’s marketing, ChatGPT is unreliable. A 2025 study, for example, found AI assistants, such as ChatGPT, misrepresent the news roughly 45% of the time. Similarly, despite marketing suggesting ChatGPT can handle financial affairs, ChatGPT has failed in meeting basic accounting standards and provided incorrect tax advice to users. The third element of the case is the public safety threat. The danger to young people in particular is illustrated by the tragic story of Adam Raine, a 16-year-old who died by suicide in April 2025 after engaging in long conversations with ChatGPT. When Adam expressed suicidal thoughts, ChatGPT responded that it “won’t try to talk you out of your feelings”. It helped Adam plan a “beautiful suicide” and even offered to write his suicide note for him. Why would a product behave this way? Because, Uthmeier argues, it was built to. OpenAI designed ChatGPT to be highly agreeable, to say “yes” roughly ten times as often as “no”, according to a Washington Post review of 47,000 conversations. This forms the fourth element of the case – commercial exploitation through sycophancy. In other words, ChatGPT optimistically parrots back users’ responses in order to to manipulate them into deeper conversations, regardless of truth or safety. But according to the lawsuit, even ordinary use carries a cost: it weakens people’s brain activity and critical thinking skills (also known as cognitive atrophy). This is the fifth element of the case. The sixth and final element is knowledge – specifically, the knowledge of Samuel Altman. According to Uthmeier, since at least 2023, OpenAI’s own documents warned that the model could coach people on committing crimes, but Altman overruled the safety staff. These six elements paint a picture of a product marketed as safe, engineered to be addictive, and known by its own makers to be dangerous – yet sold to us, anyway. Altman is at the centre of that picture. The complaint reconstructs his career and reaches for an April 2026 New Yorker investigation and testimony from the recent legal battle between Elon Musk and OpenAI to depict a man, who in Uthmeier’s telling, repeatedly chose speed over safety. That is why Uthmeier is asking the court to hold Altman personally liable for “his utter disregard for the risk to human life”. Pay for past harms Uthmeier is asking the court to declare that OpenAI broke the law, then to order the company to stop – permanently – its unlawful practices. He wants the company barred from collecting children’s data without parental consent and the safeguards that should come with it, and barred from misrepresenting or staying silent about ChatGPT’s risks. On top of the injunctions, the state is seeking civil penalties of up to US$10,000 per violation for OpenAI’s alleged wilful violation of the the Florida Deceptive and Unfair Trade Practices Act. Uthmeier said penalties could total billions of dollars. In other words: pay for the past harms and change the product going forward. In a statement to The Conversation, an OpenAI spokesperson pointed to the company’s “industry leading protections and policies” regarding user safety. In particular we built safety for minors directly into our products, including a more protective experience specifically for minors, an age prediction tool, defaulting users whose age we are not confident into our more protective experience, and giving parents tools to monitor their kids use of AI. Adding to a growing pile This lawsuit is a significant development, but it has not arrived in a vacuum. Across the US, the courts are filling with cases accusing tech companies of harming young people. In April, for example, Uthmeier launched a criminal investigation into OpenAI over the chatbot’s alleged role in a shooting at Florida State University. Some juries have started to side with the plaintiffs. In March 2026, for example, a New Mexico jury hit Meta with a US$375 million penalty in a child safety case. Days later, a jury in Los Angeles found Meta and Google liable in a landmark trial over social media addiction. This case rides the same current. But it broadens the scope by alleging Altman himself should be personally responsible. Uthmeier is demanding a trial by jury. Alexandra Andhov is the director of ALTeR (Center for Advancing Law and Technology Responsibly) at the University of Auckland. She received funding from the Independent Research Fund Denmark for the "PROFIT" Project (Gaps and Opportunities in Corporate Governance of Big Tech Companies) to research big tech companies.

<p><strong>Tool:</strong> <a href="https://tools.simonwillison.net/pasted-file-editor">Pasted File Editor</a></p> <p>I really like how you can paste a large volume of text into <a href="https://claude.ail">claude.ai</a> (or the Claude desktop/mobile apps) and it will detect it as a large paste and turn it into a file attachment instead.</p> <p>I decided to have Codex desktop <a href="https://gist.github.com/simonw/74c79119b487a5acce18b4dcc26b9f79">build me a version of that</a> as a prototype.</p> <p>You can also open files directly - including images which will be shown as thumbnails - or drag files onto the textarea.</p> <p>Tags: <a href="https://simonwillison.net/tags/javascript">javascript</a>, <a href="https://simonwillison.net/tags/tools">tools</a>, <a href="https://simonwillison.net/tags/ai-assisted-programming">ai-assisted-programming</a>, <a href="https://simonwillison.net/tags/claude">claude</a>, <a href="https://simonwillison.net/tags/codex">codex</a></p>

Yuliya Chernova / Wall Street Journal: Crunchbase: SF startups hit a record of 1,404 seed deals in 2025, while the number of NYC seed deals has dropped over the past three years, to 666 in 2025  —  'You can't ignore San Francisco right now,' says one New York venture capitalist  —  New York venture firms are broadening their geographic reach.

arXiv:2606.00082v1 Announce Type: new Abstract: Explainability of deep learning algorithms is critical for computer-vision applications with high-stake decisions. Concept bottleneck models (CBM) have recently shown promising performance to provide explainable and accurate predictions for classification problems, based on a bottleneck of high-level concepts. Existing CBM methods rely on a linear aggregation of the concept scores to compute predictions. However, a large number of concepts is often used in this linear approach, which undermines explainability and favors information leakage. In general, the underlying relation between concepts and output logits is not linear. Therefore, we introduce Hoeffding Concept Bottleneck Models (HCBM), which build on the Hoeffding functional decomposition of gradient-boosted trees to provide non-linear and sparse aggregations of concept scores, and generate compact predictions using prime implicants. HCBM are proved to be robust to interconcept leakage, and outperform standard linear CBM in practice, as shown in extensive experiments. Beyond classification, HCBM can be adapted to object detection, and we focus on a challenging case with overhead images to show the high performance of HCBM in these settings.

arXiv:2606.00083v1 Announce Type: new Abstract: Reinforcement learning relies on accurate reward functions, which are often hand-crafted or even unavailable in real-world applications, such as robotics. Recent work has explored the zero-shot reasoning capabilities of pre-trained Vision-Language Models (VLMs) as reward models. However, without careful prompt engineering, these approaches tend to produce suboptimal rewards, where false positive predictions can severely degrade downstream policy learning. In robotics, limited datasets comprising expert demonstrations are often collected to bootstrap policy learning. This scenario provides an opportunity to optimize a reward model prior policy training. We propose Demo2Reward a test-time adaptation technique to optimize the language instruction of a reward model based on a few demonstrations (3-10 trajectories) to reduce false positives while preserving true positives. Crucially, this requires no additional model training or computation resources during policy learning. We show that Demo2Reward consistently outperforms existing zero- and few-shot VLM reward models across a range of simulated robotic tasks and policy backbones. Finally, we demonstrate that Demo2Reward effectively transfers to a real-world robotic learning scenario, enabling policy learning without manually engineering a reward function.

arXiv:2606.00129v1 Announce Type: new Abstract: Large language models (LLMs) have emerged as powerful representation learners whose internal features increasingly align with human cognition. We study whether modern LLMs can serve as a lens for understanding neural representations in the human brain, focusing on emotional valence in EEG. We first build a one-dimensional valence direction, the V-axis, from modern LLMs using only nine emotion-evocative sentences. We validate it through zero-shot transfer to sentiment benchmarks and cross-model consistency across fourteen LLMs. We then show that this LLM-derived direction maps onto human neural activity. On a public EEG cohort of 123 subjects watching affective videos, a single linear projection on EEG features tracks the V-axis position of each stimulus. Moreover, 36 EEG emotion classifiers trained without exposure to the V-axis spontaneously rediscover the same direction in their internal representations, suggesting that the same valence structure emerges in both language models and human electrophysiology. Yet this convergence does not provide an effective training signal. We test twenty-five alignment strategies, including knowledge distillation, representational similarity, contrastive, and topographic losses; none improve decoding, and sixteen significantly reduce accuracy. We formalize this result as the saturation regularity: once task labels alone drive a brain-decoding network onto the target direction, additional supervision mainly distorts an already-saturated basin, while the load-bearing within-class residual receives little useful gradient. This regularity also indicates where improvement should come from: the residual subspace unreachable by supervision. Motivated by this insight, we ensemble across residual diversity rather than supervising the basin, improving balanced accuracy by 10.5% over the prior best on FACED, with the same effect replicated on SEED-V.

arXiv:2606.00132v1 Announce Type: new Abstract: While finetuning effectively adapts foundation models to specialized downstream tasks, it can degrade nontarget capabilities acquired during pretraining. Existing forgetting aware methods typically seek safer updates through specialized initialization or fixed constraints, but do not regulate the adaptation preservation trade-off during training. We propose Foundation Preserving LoRA (FoLoRA), a forgetting aware optimization framework. Guided by a first order preservation condition, FoLoRA defines a forgetting penalty over pretraining-proxy activations and a task utility over downstream task activations. It then scores update directions by task utility per unit forgetting penalty via a generalized Rayleigh quotient. The resulting spectral coordinate system enables direction wise gated Adam updates, attenuating low utility to penalty directions during training. To estimate the forgetting penalty, FoLoRA constructs pretraining proxy calibration data by sampling from the pretrained model rather than relying on a single proxy dataset. Experiments on math, code, and instruction following adaptation show that FoLoRA achieves the strongest preservation adaptation balance over baselines, improving target task performance with best aggregate preservation of non target capabilities.

arXiv:2606.00081v1 Announce Type: new Abstract: Distributed Acoustic Sensing (DAS) enables large-scale monitoring through optical fibers, but its high dimensionality and complex spatio-temporal patterns make event classification demanding. Existing deep learning approaches-CNNs, recurrent models, and Transformer variants-either fail to capture long-range dependencies or require processing raw DAS matrices at prohibitive cost. We propose DAStatFormer, a hybrid multibranch Transformer that combines compact multidomain statistical features with Gated Transformer Networks. Instead of raw signals, we extract 24 ANOVA-selected attributes per channel from the temporal, waveform, and spectral domains, reducing data size by orders of magnitude while preserving discriminative information. Each domain is processed via dedicated step-wise and channel-wise attention branches, fused by an adaptive gating mechanism. Experiments on the open $\Phi$-OTDR benchmark and a real-scenario DAS dataset show that DAS-tatFormer achieves up to 99.4% accuracy and near-perfect real-world performance, while using significantly fewer parameters and lower inference cost than models such as DASFormer and DeepViT. These results demonstrate its suitability for scalable, real-time DAS-based monitoring. We release our code at https://github.com/MichelD-git/DAStatFormer

arXiv:2606.00079v1 Announce Type: new Abstract: Mixture-of-Experts (MoE) large language models reduce per-token computation through sparse expert activation, but their deployment remains memory-intensive because all expert weights must be kept resident in memory. Existing MoE compression methods struggle in the ultra-low-bit regime: pruning irreversibly removes model capacity, while coarse-grained quantization fails to allocate bits according to heterogeneous expert and weight-direction importance. We propose BitsMoE, a spectral-energy-guided bit-allocation framework for MoE LLM quantization. BitsMoE decomposes each MoE layer by SVD into a shared basis and expert-specific spectral factors, retaining the shared basis without quantization to preserve common cross-expert structure and using the expert-specific factors as fine-grained quantization units. To determine the bit-width of each unit, BitsMoE formulates spectrum-wise mixed-precision quantization as an activation-aware reconstruction surrogate and solves an integer linear program that minimizes estimated reconstruction loss under a fixed bit budget. Experiments across multiple MoE LLMs show that BitsMoE substantially reduces downstream task accuracy degradation in ultra-low-bit regimes. Under 2-bit quantization on Qwen3-30B-A3B-Base, BitsMoE accelerates quantization by 12.3$\times$, improves average accuracy by 27.83 percentage points, and increases decoding speed by 1.76$\times$ over GPTQ. Our model and code are publicly available at https://github.com/zjiayu064/BitsMoE.

arXiv:2606.00094v1 Announce Type: new Abstract: Image generative models aim to sample data points from the underlying data manifold, a task that requires learning and decoding a dense, low-dimensional, and compact parameterization space. To achieve this, we propose the Data Manifold-aware Image diffusioN moDel (MIND), a novel framework that explicitly models manifold geometry by integrating discrete patch tokenization into the score function of a continuous diffusion model. This approach successfully leverages both the structural quantification capabilities of discrete tokens and the parallel generation flexibility of continuous diffusion. Moreover, we enable end-to-end differentiable training via a novel soft top-$k$ aggregation mechanism and introduce dual-branch high-frequency feature embedding layers to alleviate the spectral bias of transformer backbones on low-dimensional inputs. Furthermore, for inference, we design a multi-stage transition sampling scheme that dynamically adjusts the sampling scheme based on timestep. Extensive experiments on ImageNet 256$\times$256 demonstrate the effectiveness of MIND. After 80-epoch training, our base model achieves an FID of 22.73 without guidance, nearly halving the 43.47 FID of the vanilla DiT-B/2 baseline. The proposed method reduces FID by 15.95 and 9.06 on average compared with the baselines DiT and SiT, respectively. For image generation on ImageNet-256$\times$256 with guidance, the proposed MIND-B with only 130M parameters achieves an FID of 2.06, superpassing the LlamaGen-3B with 3.1B parameters. The proposed MIND-XL with 715M parameters further reduces the FID to 1.95. Our MIND introduces a fresh perspective on diffusion-based image generation, paving the way for future research and innovation in this community. The code will be publicly available.

arXiv:2606.00024v1 Announce Type: new Abstract: Long-context decoding in Large Language Models (LLMs) is severely constrained by the memory bandwidth required to fetch the extensive Key-Value (KV) cache. Most existing KV management methods rely on key-only pruning before decoding, despite the evidence that attention outputs depend jointly on keys and values, as incorporating values in their methods incurs prohibitive additional overhead. In this paper, we propose Attention Run-time Termination (ART), a lightweight run-time mechanism that tracks accumulated attention outputs during kernel execution and terminates subsequent KV block accesses once further contributions become negligible. This design makes ART orthogonal to existing key-based KV cache management methods, enabling seamless integration with them. Experiments on LongBench benchmarks show that ART achieves 20% higher generation throughput in large batch size than state-of-the-art baseline while maintaining comparable accuracy.

arXiv:2606.00023v1 Announce Type: new Abstract: The rapid development of Language Diffusion Models (LDMs) challenges the dominant position of auto-regressive competitors in language processing. However, their flexible, any-order decoding strategies not only enable fast decoding speed but also potentially bring new trustworthiness challenges. To better understand the risks behind their pipelines, we introduce a comprehensive trustworthiness benchmark tailored to LDMs (TrustLDM), evaluating safety, privacy, and fairness across different LDM architectures with multiple categories of static post contexts. Our empirical results show that although LDMs generally exhibit strong trustworthiness with only the user prompts, their alignment behavior degrades noticeably when the malicious post contexts are attached to the masked responses. We further observe that longer contexts do not necessarily induce stronger effects, and both decoding order and generation length affect the evaluation outcomes. Finally, we propose TrustLDM-Auto, an automatic evaluation framework that leverages LDM decoding flexibility to systematically identify vulnerable configurations, revealing substantial trustworthiness weaknesses across all evaluated models and dimensions. Our work may potentially help the community build more trustworthy LDMs. Our code is available at https://github.com/PKU-ML/TrustLDM.

arXiv:2606.00022v1 Announce Type: new Abstract: Humor generation remains difficult not only because producing fluent, novel jokes is hard, but because "funny" is audience-dependent and supervision is noisy -- preferences vary with audience, context, and culture, and annotator agreement is often low. In this paper, we describe our system for the SemEval-2026 Task-1 (MWAHAHA), which focuses on humor generation under explicit constraints. The task evaluates submitted systems via human preference judgments in 1-on-1 arena-style comparisons. We adopt a "generate-many -> select-best" strategy. First, we generate a diverse pool of candidates per instance using multi-step prompting, model ensembling, and diversity-oriented decoding. Second, we select outputs using a preference model that approximates a "reader" by learning from human comparisons rather than absolute funniness scores. To support this approach, we release 2.5K human pairwise judgments collected through the Humor Arena prototype. We further propose an interpretable pipeline that converts labeled comparisons into a preference model. Across three preference datasets, our models consistently outperform baselines and show stronger cross-domain transfer. Finally, we apply the learned preference model to rank candidates for the MWAHAHA setting and release intermediate artifacts (candidate pools and rankings) to facilitate follow-up work. Our system ranked 1st in the English and Chinese subtasks of MWAHAHA and 2nd in the Spanish subtask.

arXiv:2606.00021v1 Announce Type: new Abstract: Speculative Decoding (SD) accelerates Large Language Model (LLM) inference by employing a lightweight draft model to propose candidate tokens, which are verified in parallel by the target model, without compromising generation quality. While Retrieval-based Speculative Decoding (RSD) is favored for its plug-and-play versatility, its potential is impeded by rigid lexical dependencies, rendering both retrieval and verification brittle to surface-level variations. To address this, we propose SENSE (Semantic Embedding Navigation with Soft-gated Evaluation). By anchoring retrieval on the hidden states of the target model, SENSE establishes robust semantic alignment, which empowers the Soft-gated Evaluation module to validate semantic equivalence rather than surface forms. To ensure rigorous benchmarking, we deconstruct existing methods into atomic primitives within a unified framework, facilitating granular, component-level comparison. Extensive experiments across diverse domains demonstrate that SENSE outperforms multiple baselines on the LLaMA and Qwen families, attaining up to 4.09 mean acceptance length and 3.26x speedup, while preserving generation quality. Our code will be released upon publication.

arXiv:2606.00020v1 Announce Type: new Abstract: Large Language Model (LLM) based Chinese Grammatical Error Correction (CGEC) systems face two critical challenges: general-purpose models lack specialized linguistic priors for subtle grammatical distinctions, and Supervised Fine-Tuning (SFT) with Maximum Likelihood Estimation fails to optimize for precision-focused metrics, leading to systematic over-correction. We propose CSRP, a three-stage framework that progressively builds correction capability through Continual Pre-training (CPT) on 5.9M balanced samples to internalize domain knowledge, Chain-of-Thought SFT with explicit error reasoning for diagnostic transparency, and Group Relative Policy Optimization with a novel Efficiency-Aware Reward that explicitly penalizes unnecessary edits. On the NACGEC benchmark, CSRP achieves state-of-the-art performance with 50.99 $F_{0.5}$ and 57.17 precision, substantially outperforming previous best results while effectively mitigating the over-correction bias inherent in MLE-trained models. Our method also advances CSCD spelling correction to 59.61 F1, surpassing GPT-4 by 5.20 points. Comprehensive ablation studies demonstrate that the RL alignment stage contributes a 8\% relative gain over the SFT baseline, and that this gain is orthogonal to the contribution of large-scale CPT, validating that explicit optimization for edit efficiency is essential for high-quality grammatical error correction. Our code is available at https://github.com/TW-NLP/ChineseErrorCorrector.

arXiv:2606.00016v1 Announce Type: new Abstract: Detecting AI-generated text is becoming increasingly challenging as modern language models approach human-level fluency and can evade detectors that rely on surface statistics or likelihood-based signals. We propose \textsc{AEyeDE}, an attribution-driven approach to human-AI authorship detection that leverages model attention as a discriminative signal. Specifically, we extract attention-based attribution matrices for both human- and AI-generated text using a \emph{proxy} Transformer model with white-box access and train a lightweight Convolutional Neural Network to learn representations from these attribution maps. Across encoder-decoder translation settings, our method consistently outperforms a text-only baseline. In decoder-only settings, it performs strongly in generator-specific detection, remains competitive on standard benchmarks, and shows robustness under cross-dataset transfer and alternative-spelling perturbations. We further show that attention maps exhibit recurring local structures whose relative frequencies differ consistently between human- and AI-generated text across datasets and proxy models. These findings suggest that attention-based attribution maps provide a complementary and interpretable signal for AI-generated text detection. We will make the code publicly available to support future research.

arXiv:2606.00014v1 Announce Type: new Abstract: Although studies have demonstrated that Large Language Models (LLMs) can perform well on Out-of-Distribution (OOD) tasks, their advantage tends to diminish as the distribution shift becomes more severe. Consequently, researchers aim to retrieve distributionally similar and informative demonstrations from the available source domain to boost the inference capabilities of LLMs. However, in practical scenarios where the target domain is inaccessible, evaluating the unknown distribution is challenging, which indirectly impacts the quality of the selected demonstrations. To address this problem, we propose \textbf{DOPA}, a demonstration search framework that incorporates an OOD proxy to approximate the inaccessible target domain and guide the retrieval process. Building on proxy-based evaluation, DOPA further introduces a Mahalanobis distance-based global diversity constraint to ensure sufficient diversity among the retrieved demonstrations. Experimental results on multiple LLMs and tasks demonstrate that DOPA effectively enhances robustness in OOD settings\footnote{https://github.com/bort64/ood\_code}.