AI News Today

arXiv:2607.20811v1 Announce Type: new Abstract: In spite of the fundamental role of neural networks in contemporary machine learning research, our understanding of the computational complexity of optimally training neural networks remains incomplete even when dealing with the simplest kinds of activation functions. Indeed, while there has been a number of very recent results that establish ever-tighter lower bounds for the problem under linear and ReLU activation functions, less progress has bee

arXiv:2607.20822v1 Announce Type: new Abstract: Motivated by reinforcement learning in harsh environments, we consider the problem of learning an optimal policy subject to adversarially corrupted feedback. Specifically, at each time-step, an adversary can perturb both the reward and state observations of the learner following the Huber contamination model. To defend against such data corruption, we propose {{\texttt{BR-Async-Q}}}: a novel, epoch-based, robust \(Q\)-learning algorithm built upon

arXiv:2607.20425v1 Announce Type: new Abstract: What makes writing "good" remains a persistent question in literary studies and computational linguistics. We present a two-study investigation of how reasoning-enabled LLMs evaluate literary quality. In Study 1, we construct a benchmark of 30 real texts spanning six quality tiers, from canonical literature to anonymous forum posts, and extract the model's implicit theory of quality from its reasoning traces. Across five DeepSeek replications, the

arXiv:2607.20426v1 Announce Type: new Abstract: Existing LLM hallucination mitigation methods, including prompt engineering and model optimization, either hardly alter models'internal knowledge or have poor cross-domain generalization. Contrastive decoding mitigates hallucinations by using layer-wise differences in LLMs. However, prior studies only explore transformer-based models (e.g., GPT), ignoring other effective frameworks like mixture-of-experts (MoE) models. Since MoE alters the traditio

arXiv:2607.20428v1 Announce Type: new Abstract: This study evaluated a retrieval-augmented, multi-agent large language model (LLM)-driven, human-in-the-loop framework for detecting cutaneous immune-related adverse events (cirAEs) from clinical notes. Compared with unassisted manual review, the LLM-assisted workflow improved accuracy (F1 = 0.88 vs 0.77), inter-rater agreement measured by Cohen's kappa (kappa = 0.82 vs 0.50), and reduced average review time by approximately half. This framework pi

arXiv:2607.20429v1 Announce Type: new Abstract: Large language models are increasingly used to simulate diverse human opinions in open-ended tasks such as synthetic surveys, focus group modeling, and public opinion prediction. However, LLM outputs exhibit systematic opinion homogenization. Practitioners have explored various interventions to increase diversity, but the landscape remains fragmented: different methods are evaluated in isolation with incomparable metrics, and in practice they are t

arXiv:2607.20430v1 Announce Type: new Abstract: We present LLM-INSTRUCT, the winning system for the UZH Shared Task at ArgMining 2026 on paragraph-level argument mining in UN and UNESCO resolutions. The task requires paragraph-type classification, prediction of a subset of 141 official tags, and directed relation prediction under a strict JSON schema setting using only open-weight models up to 8B parameters. We frame the task as constrained structured prediction. The system first narrows the can

arXiv:2607.20431v1 Announce Type: new Abstract: Materials science literature analysis requires simultaneous attention to composition, processing, characterization, and property relationships, yet conventional retrieval-augmented generation pipelines struggle to reconcile heterogeneous tasks within a single retrieve-then-generate architecture. Here we present AlphaAgent, a skill-driven agent framework that decouples retrieval-based question answering from paper-level report generation through exp

arXiv:2607.20432v1 Announce Type: new Abstract: Recent advances in large language models and their widespread adoption have prompted claims that natural language could entirely replace formal languages, such as programming languages for software design. In this position paper, we argue that this perspective overlooks fundamental linguistic properties of natural language, specifically that it is optimized for underspecification in open-ended contexts. We introduce a formal framework centered on *

arXiv:2607.20433v1 Announce Type: new Abstract: While language models remain frozen at their training state, the world evolves continuously. Knowledge editing has emerged as a key alternative to full retraining, but its deployment is bottlenecked by the erosion of core capabilities: mathematical and programmatic reasoning collapse while encyclopedic recall remains intact. We trace this asymmetric degradation to a distributional mismatch. Covariance-based editors preserve only the subspaces spann

arXiv:2607.20435v1 Announce Type: new Abstract: Open-source LLMs (OSMs)arereaching near state-of-the-art performance, prompting prior works to trace the text they generate by embedding text watermarking algorithms directly into their weights. Yet, OSMs are subject to post-training modifications, which has been shown to remove the watermark. Model merging in particular, a prominent method used for combining expert knowledge and preventing catastrophic forgetting, strongly removes such OSM waterma

arXiv:2607.20436v1 Announce Type: new Abstract: Safety evaluations often assume that behavior observed during testing reflects behavior in ordinary use, but fine-tuning can break this assumption. A checkpoint can appear fixed under evaluation-style prompts while the same behavior persists under ordinary-use prompts. Output scores reveal this mismatch but do not locate it. We investigate whether the distinction is encoded in a stable internal site and introduce an approach that fits a paired acti

arXiv:2607.20438v1 Announce Type: new Abstract: Preference-based post-training is usually understood through endpoint behavior, yet the learned update that produces this behavior remains largely opaque. We study RLHF and related preference optimization through the spectral structure of their induced parameter updates. By decomposing effective LoRA updates and reloading their spectral components as plug-in modules, we turn preference-induced updates into objects that can be isolated, recomposed,

arXiv:2607.20440v1 Announce Type: new Abstract: Proprietary large language models (LLMs) entail substantial intellectual and financial investment, making them valuable intellectual property (IP). However, even when deployed via black-box APIs, these models remain vulnerable to unauthorized knowledge distillation, which allows adversaries to cheaply extract and replicate model capabilities. To address this issue, anti-distillation (AD) has been proposed to generate defensive outputs that hinder d

arXiv:2607.20441v1 Announce Type: new Abstract: Every information ecosystem produces beliefs that shape strategic decisions. Both human analysts and AI systems inherit the blind spots of their information sources. We show that LLMs, combined with prediction markets, function as a calibrated instrument for measuring how far ecosystem-induced beliefs deviate from an external reference: LLMs extract the beliefs a text corpus implies, and prediction market price trajectories, anchored at resolution

arXiv:2607.20442v1 Announce Type: new Abstract: We release Naver-News-KO, a Korean news summarization dataset of 27,400 (document, summary) pairs collected from Naver News over a ten-day window in July 2022 across two categories (Economy and IT/Science; 77/23 split), with train/validation/test partitions of 22,194 / 2,466 / 2,740 and a mean per-record document-to-summary character-compression ratio of 6.03x. The dataset has been publicly hosted on the Hugging Face Hub since January 2023 and, as

arXiv:2607.20444v1 Announce Type: new Abstract: Large language models (LLMs) can produce deceptive responses: outputs that mislead users in service of a contextually or experimentally induced goal. Yet it remains unclear how confidently models deceive and whether higher confidence makes deceptive responses more persuasive to end users. In this paper, we study these basic questions in various models and different deception datasets. We provide a comprehensive study measuring confidence through bo

arXiv:2607.20446v1 Announce Type: new Abstract: As large language models (LLMs) are increasingly used by students to generate natural language responses and program code, there is growing interest in whether LLMs themselves can be used to distinguish AI-generated work from human-authored submissions. In this paper, we investigate the extent to which LLMs can detect their own generated content across multiple educational task types, including programming exercises, reflective writing, and short-a

arXiv:2607.20447v1 Announce Type: new Abstract: This paper describes our system for the EEUCA 2026 Shared Task on toxicity classification in gaming chat. We implement a three-stage pipeline combining an ensemble of two compact transformers (DeBERTa-v3-base, 184M; XLM-RoBERTa-base, 278M) with a Linguistically-Informed Mediator (LIM) that resolves inter-model disagreements through corpus-backed lexical normalization, class-conditional unigram scoring, multilingual profanity detection, and agentive

arXiv:2607.20448v1 Announce Type: new Abstract: We introduce Domyn-Small, a 10-billion-parameter open-weight reasoning language model released under the MIT license. Domyn-Small is the product of an initial pre-training phase on 9 trillion tokens multilingual data, followed by a post-training pipeline for reasoning, instruction following, and context extension. For the latter, we performed a Continued Pre-Training (CPT) phase that doubles the native context window to 32K tokens, followed by SFT

arXiv:2607.20450v1 Announce Type: new Abstract: This paper presents two systems for the GameTox Shared Task at the Workshop on EEUCA at ACL 2026, which requires classifying World of Tanks chat utterances into six fine-grained toxic intent categories (Labels 0-5). Severe class imbalance, domain-specific multilingual slang, and extremely scarce data for rare categories such as Threats (Label 4, 60 samples) and Extremism (Label 5, 24 samples) make this a challenging classification problem. Our prim

arXiv:2607.20451v1 Announce Type: new Abstract: Semantic Field Theory (SFT) has developed from a philosophical critique of strong anti-formalist readings of language games into a proposed computational model class for lexical semantics, higher order composition, and stabilized interpretation. This paper reconstructs that evolution and gives SFT a sharper mathematical core suitable for independent evaluation in computational linguistics and representation learning. The central proposal is that a

<p>When an AI agent leaks data, it may not look like a breach at first. It may look like a normal tool call, a helpful API request, or a browser fetch that quietly sends the wrong payload to the wrong place.</p> <p>That is the uncomfortable part for builders: prompt safety can warn you about intent, but only the network boundary can stop bytes from leaving.</p> <p>If your product lets agents call APIs, browse pages, use MCP tools, fetch files, or run long workflows, you need a simple rule: agent

Hey HN, I built GlossaryPage after noticing companies have long-tail content opportunities that never get created because they fall between blog posts and documentation. The primary use case is marketing teams building glossaries and learning hubs around these topics. The same workflow also works for support teams creating help centers. GlossaryPage helps teams identify unanswered topics, generate drafts with AI, review content, and publish structured knowledge pages. Here is a demo site: https:

<pre style='white-space:pre-wrap;width:81ex'>feat(browser): add `openclaw browser batch` CLI subcommand (#111457) * feat(browser): add `openclaw browser batch` CLI subcommand Expose the existing `act:batch` runtime through a dedicated CLI subcommand so users and scripts can run nested act requests in one call without going through the agent tool. - `--actions <json>` for inline JSON, `--actions-file <path>` for file input, `--actions-file -` for stdin (1MB cap to bound a runaway pipe) - `--conti

<p>If you've been anywhere near AI job postings lately, you've seen the title: <strong>Forward-Deployed Engineer.</strong> Sometimes it's "Deployment Engineer" or "Solutions Engineer" or "Applied AI Engineer." Sarvam is hiring more than a hundred of them. Palantir built a large part of its business on them. OpenAI, Anthropic, and a long tail of AI startups are all competing for the same people.</p> <p>And yet, if you ask five engineers what an FDE <em>is</em>, you'll get five different answers.

Article URL: https://bohay.dev/ Comments URL: https://news.ycombinator.com/item?id=49030938 Points: 1 # Comments: 0

I noticed that recently some stories are submitted more than once, and the duplicates make it to the front pages. I thought HN detects this and redirects similar submissions to the first one. Examples: Show HN: Whetuu – a zero-config cross-shell prompt written in Zig (https://github.com/yamafaktory/whetuu) https://news.ycombinator.com/item?id=48991361 https://news.ycombinator.com/item?id=48981557 Stop pretending billionaires built the future (https://www.elysian.press/p/billionaires-didnt-create

<pre style='white-space:pre-wrap;width:81ex'>fix(auth-profiles): preserve secret refs and OAuth fields during doctor auth migration (#97881) Both legacy auth migration writers rebuilt each credential from a hardcoded field subset, dropping keyRef/tokenRef and the OAuth clientId/idToken/ chatgptPlanType, even though the canonical reader already preserves them. applyLegacyAuthStore now keeps the already-parsed credential as-is, and coerceLegacyFlatCredential delegates field extraction to the share

<h1> Bio-Tuning Glasses: Building an Invisible Biofeedback Interface with Edge AI and Adaptive Optics </h1> <p>What if smart glasses didn't constantly tell you how healthy—or unhealthy—you are?</p> <p>No step counts.<br> No stress notifications.<br> No endless dashboards.<br> No digital reminders telling you to "sit straight" or "go to sleep."</p> <p>Instead, imagine a wearable device that quietly adapts the environment around you based on your physiological state.</p> <p>This is the idea behind

<p>Controller A records revocation epoch 12. Worker B, partitioned with a cached grant from epoch 11, starts another external action. The database is correct and the system is unsafe. Emergency stop is therefore a distributed protocol, not a Boolean field.</p> <h2> What is verified </h2> <p>In its July 21 disclosure, OpenAI says an internal benchmark used models with reduced cyber refusals and that a combination of models compromised Hugging Face infrastructure. The primary source is <a href="ht

<p>A reviewer approves “update dependencies,” but the system later interprets that as publishing a package. The human was present; meaningful approval was not. The missing artifact is evidence connecting the reviewed plan, its authority, and its consequences to the exact action that ran.</p> <h2> What is verified </h2> <p>According to OpenAI's July 21 disclosure, a combination of models operating in an internal benchmark with reduced cyber refusals compromised Hugging Face infrastructure. The pr

<p><em>Turning entire novels into narrated, lip-synced, motion video — locally, on a single 16GB GPU, with FLUX, Wan2.2, PuLID, MuseTalk, and ComfyUI.</em></p> <p>There's an old line: a picture is worth a thousand words. I'd extend it — a video is worth a thousand pictures. So I spent the last few months testing that idea the hard way: I built a pipeline that turns an entire novel into a narrated, lip-synced, motion video. Feed it <em>Pride and Prejudice</em>, or a 400-year-old tale from <em>Str

Article URL: https://tuziyo.com/ Comments URL: https://news.ycombinator.com/item?id=49030797 Points: 1 # Comments: 0

Hi HN! I built PDFly — a PDF toolkit that runs entirely client-side. No file uploads, no server processing, no account needed. It has 56 tools including merge, split, compress, convert, redact, compare, and an AI chat feature to ask questions about your PDF content. Everything runs in your browser using [mention your tech stack if relevant, e.g., WASM/JS libraries]. Would love feedback on the tools, UX, or anything else! Comments URL: https://news.ycombinator.com/item?id=49030778 Points: 2 # Com

<pre style='white-space:pre-wrap;width:81ex'>refactor(agents): close embedded attempt terminal outcomes (#113201) * refactor(agents): close embedded attempt outcomes * test(agents): align trajectory terminal fixture * refactor(codex): share attempt failure type * fix(codex): retain projector result contract * test(agents): keep terminal projections test-only * chore(plugin-sdk): refresh attempt terminal baseline</pre>

Article URL: https://www.meyo.life/skill Comments URL: https://news.ycombinator.com/item?id=49030664 Points: 2 # Comments: 0

https://blog.seaticket.ai/how-to-resolve-software-deployment-issues-with-seaticket-8cff551ab44e Comments URL: https://news.ycombinator.com/item?id=49030661 Points: 1 # Comments: 0

We spoke with several cybersecurity researchers, who look for unknown vulnerabilities and develop tools to exploit them, about how OpenAI’s and Anthropic’s guardrails affect their work.

I wrote an interactive introduction to IEEE-754 floating point and would love some feedback. My main issue with existing treatments of this topic (and internet tutorials, broadly) is that they ask too little of the reader. I wanted to structure this around a sequence of questions for you to answer, using interactive widgets and optionally an LLM tutor/grader. Basically: you progress in the tutorial by demonstrating your understanding. The feedback I'm looking for is mostly around structure. Does