AI News Today

arXiv:2606.22478v1 Announce Type: new Abstract: Multilingual Language Models like mBERT are widely used for low-resource NLP, yet their adaptation to morphologically inconsistent languages such as Roman Urdu remains underexplored. Roman Urdu spelling variation causes severe sub-word fragmentation, averaging 1.50 sub-words per token. We propose \textit{ROMEVA} (Roman Urdu Embedding-preserving Vocabulary Adaptation), which combines sub-word-average initialization and a PCA-guided anchor loss to st

arXiv:2606.22511v1 Announce Type: new Abstract: In open-ended generation, LLMs frequently fall into the "likelihood trap", marked by repetitive degeneration and vocabulary dullness, creating a discrepancy between machine-generated and human-written text. While post-hoc tail truncation (e.g., Top-$p$, Min-$p$) avoids sampling from the unreliable tail, it can over-sample from the uncalibrated head and misalign generation with human lexical preferences; fixed scalar repetition penalties likewise ig

arXiv:2606.22565v1 Announce Type: new Abstract: Chain-of-Thought (CoT) has become a standard method for improving reasoning capabilities in large language models (LLMs) by eliciting step-by-step thinking, but its effectiveness in multimodal tasks remains unclear. In this paper, we aim to systematically investigate the key question: What can multimodal Chain-of-Thought reasoning do, and where and why does it fall short? To this end, we evaluate 12 multimodal tasks across perception and reasoning

arXiv:2606.22570v1 Announce Type: new Abstract: Reinforcement Learning from Verifiable Rewards (RLVR) has emerged as a promising framework for enhancing the reasoning ability of large language models. However, much of the existing work is guided by heuristic intuition, leading to divergent algorithmic choices, even contradictory ones that nevertheless report empirical gains. To better understand this phenomenon, we conduct a theoretical analysis of RLVR updates. Our study reveals that difference

arXiv:2606.22578v1 Announce Type: new Abstract: We extend our prior work on Text2DSL automatic generation of domain-specific language (DSL) code from natural language descriptions along two complementary axes. First, we replace prompt-only synthetic generation with context-aware distillation, in which a teacher large language model (DeepSeek-V4-Flash) operates under an explicitly defined structured context comprising a BNF grammar, an API specification, and a closed identifier vocabulary; the re

arXiv:2606.22606v1 Announce Type: new Abstract: Large language models (LLMs) achieve strong relation extraction (RE), but their computational demands and reliance on proprietary APIs limit deployment in resource-constrained or privacy-sensitive settings. We investigate how far small language models (SLMs) can close this gap across general-domain and literary text. We evaluate five models from 360M to 3B parameters under three domain-composition regimes and two prompt-conditioned tuning styles (3

arXiv:2606.22627v1 Announce Type: new Abstract: Knowledge editing aims to efficiently update factual information in Large Language Models (LLMs) without full retraining. However, existing methods still suffer from performance degradation in batch knowledge editing. We identify that semantic representation entanglement, such as overlapping concepts and shared syntactic patterns, accumulates interference in the representation space and reduces editing precision. To bridge this gap, in this paper,

arXiv:2606.22722v1 Announce Type: new Abstract: Encoder-only transformer models remain essential for production NLP pipelines. We introduce moBERTo, a Portuguese adaptation of ModernBERT obtained through continued pretraining of the ModernBERT-base checkpoint on 60 billion tokens (5 epochs over a 12-billion-token corpus curated from FineWeb2 and filtered with educational and STEM classifiers). We preserve the original architecture, including rotary positional embeddings, alternating local-global

arXiv:2606.22723v1 Announce Type: new Abstract: Although Large Language Models (LLMs) excel in many tasks, their assessment in Portuguese has received less attention, particularly for open-ended, discursive tasks that demand deeper reasoning and generation capabilities. While the original BLUEX benchmark addressed the scarcity of Portuguese evaluation datasets through multiple-choice questions from Brazilian university entrance exams, it did not cover the more challenging second-phase examinatio

arXiv:2606.22728v1 Announce Type: new Abstract: The trustworthiness of a retrieval-augmented generation (RAG) system depends on more than the answer it returns, yet many black-box uncertainty methods still read agreement among sampled answers as confidence. That inference fails when repeated samples condition on the same defective retrieval state. The state may be empty, with the model falling back on parametric memory, or populated by a coherent but wrong neighbourhood. In either case, the answ

arXiv:2606.22745v1 Announce Type: new Abstract: This study investigates sentiment polarity biases, specifically, differences in how accurately AI models classify positive versus negative reviews across languages and model architectures. Large language models show a negative bias in French and are more accurate on negative reviews, while encoder models exhibit positive bias in Japanese, missing negative reviews that use indirect criticism. These language-specific polarity biases have implications

arXiv:2606.22748v1 Announce Type: new Abstract: Some professional authors are beginning to use AI tools to help produce their fiction writing. Are readers using AI to generate fiction, too? This paper examines how large language models are reshaping the production and consumption of fiction by enabling new forms of participation in narrative generation. Drawing on over 500,000 anonymized, English-language ChatGPT-user conversations (arXiv:2405.01470), we find that more than one third of the conv

arXiv:2606.22771v1 Announce Type: new Abstract: Understanding moral values in social media text offers insight into moral judgement formation, and supervised NLP models trained on crowdsourced data have achieved strong classification performance. However, most approaches simplify the problem by aggregating multiple annotators' labels into a single "ground truth", overlooking the inherent subjectivity of the task. In practice, there are disagreements between annotators caused by personal viewpoin

arXiv:2606.22811v1 Announce Type: new Abstract: Classical TTS systems typically rely on rigid input formats and predefined metadata slots, limiting their ability to fulfill flexible user requirements. This paper introduces Bagpiper-TTS, a universal speech synthesis system that deals with diverse natural language user requests. Given a natural language prompt, Bagpiper-TTS first reasons over the users' intent to derive a rich caption, i.e., a comprehensive textual blueprint encompassing both tran

arXiv:2606.22841v1 Announce Type: new Abstract: As Large Language Models (LLMs) achieve widespread integration across diverse linguistic landscapes, ensuring their safety and alignment with regional normative values remains a critical challenge. Current safety mechanisms are predominantly optimized for English-centric frameworks, often failing to capture the unique socio-cultural sensitivities and localized categories of harm inherent to the Indic region. To address this gap, we introduce IndicG

arXiv:2606.22877v1 Announce Type: new Abstract: LLM agents increasingly act as personal assistants that must remember a user's profile over months: who they are (attributes), what they routinely do (habits), and what they prefer (preferences), and keep it updated as jobs, routines, and tastes drift. Existing benchmarks evaluate this "memory" ability through short, simplified interactions, missing three core properties of real behavior: the profile is heterogeneous, with attributes, habits, and p

arXiv:2606.22942v1 Announce Type: new Abstract: Large language models (LLMs) achieve strong performance across many tasks, but their high computational cost limits deployment in resource-constrained environments. Knowledge Distillation (KD) offers a practical solution by transferring knowledge from a teacher model of a larger size to a smaller student model. While prior work has mainly examined task-specific or small-scale settings, the post-training stage for building general instruction-follow

arXiv:2606.22977v1 Announce Type: new Abstract: Statistical analysis is a broad, complex field requiring both domain knowledge and tool proficiency. While prior work has evaluated large language models (LLMs) in this domain, existing benchmarks remain limited in scope and format. To bridge this gap, we introduce StatABench (Statistical AnalysisBenchmark), a benchmark designed to systematically assess LLMs' statistical analysis capabilities. StatABench comprises two complementary components: Stat

arXiv:2606.22992v1 Announce Type: new Abstract: Knowledge graphs (KGs) are increasingly used as structured context for Large Language Models (LLMs), but industrial KG-RAG systems often need to integrate public and domain-specific KGs constructed from heterogeneous databases. This integration relies on Entity Alignment (EA), where lexical matching alone is insufficient under predicate-name variation and incomplete local neighborhoods. We address EA for KG integration by constructing a pairwise EA

OpenAI on Monday said it's releasing an improved version of its GPT‑5.5‑Cyber model to trusted defenders as part of the Daybreak initiative the artificial intelligence (AI) company announced last month. Calling GPT‑5.5‑Cyber its "strongest model yet for finding and helping patch software vulnerabilities," OpenAI said the model can "sustain deeper analysis across large codebases" to

Trying to figure out if there are already companies that have build firewall like products for AI models. Assuming everyone will now start hosting open source models to control their destiny, I wonder if this is a large market for startups to go after. Comments URL: https://news.ycombinator.com/item?id=48640044 Points: 1 # Comments: 4

Article URL: https://loomstack.co/ Comments URL: https://news.ycombinator.com/item?id=48639993 Points: 1 # Comments: 0

Article URL: https://hey.paris/posts/ai-data-centres-tasmania/ Comments URL: https://news.ycombinator.com/item?id=48639963 Points: 3 # Comments: 1

<p>I had a long Google AI Studio (Gemini) session that I wanted to keep. I selected the conversation in the browser, copied it, and pasted it into a text file.</p> <p>File size: a few hundred KB. "OK, that's safe."</p> <p>Later, I exported the same session as JSON. File size: a few MB.</p> <p>More than half of the data had silently disappeared.</p> <h2> What was missing </h2> <p>I checked what the manual copy had dropped.</p> <h3> The system prompt </h3> <p>The instruction I had originally given

<p>LinkedIn automation has a trust problem.<br> Not with users — with LinkedIn itself.<br> Most automation tools treat LinkedIn's API like an obstacle to route around. They send at fixed intervals, ignore behavioral limits, and optimize purely for volume. The result: accounts flagged within weeks, connection limits imposed, and in the worst cases — permanent bans.<br> When we built SendCopy.ai, we approached this differently. Here is the technical breakdown of how we built LinkedIn outreach auto

<p><a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fwn2kpejdc8tzfftisfdc.png" class="article-body-image-wrapper"><img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fwn2kpejdc8tzfftisfdc.png" alt=" " width="800" height="4

<pre style='white-space:pre-wrap;width:81ex'>fix(agents): align OpenRouter model scan body cap</pre>

<pre style='white-space:pre-wrap;width:81ex'>fix(agents): bound Google prompt cache response reads (#95417) The Google embedded-agent prompt-cache helpers parsed cachedContents metadata with an unbounded `await response.json()` in both createGooglePromptCache and updateGooglePromptCacheTtl. A buggy or hostile Generative Language endpoint returning a 200 with a large or never-ending body (especially with no Content-Length) would be fully buffered into memory before parsing, with the existing canc

Comments

<blockquote> <p><strong>Note:</strong> This post contains architecture diagrams. For the fully rendered version visit the <a href="https://blogs.sourishchakraborty.com/open-source-ai-coding-ide-architecture" rel="noopener noreferrer">original post</a>.</p> </blockquote> <p>AI coding assistants have fundamentally changed how developers write software. The best ones do one thing well: collapse the distance between the developer's intent and the code that expresses it. But under the hood, they are

<p>Cross-posted from my blog. Canonical version: <a href="https://launchsoloai.com/insights/runcap-vs-langfuse-vs-litellm-ai-cost-control" rel="noopener noreferrer">https://launchsoloai.com/insights/runcap-vs-langfuse-vs-litellm-ai-cost-control</a></p> <p>You let a coding agent loose on a task. It loops. It re-reads the same files, re-summarizes the same context, retries the same failing call. Forty minutes later you check the provider dashboard and the run cost more than the feature was worth.

<pre style='white-space:pre-wrap;width:81ex'>fix(agents): bound OpenRouter model-scan catalog success body (#95418) The OpenRouter /models catalog read in fetchOpenRouterModels hardened only the error/early-return path (dbd5689 cancels the body when res.bodyUsed is false), but the success branch still buffered the whole body with an unbounded `await res.json()`. The response is a provider-controlled, runtime-fetched body, so a faulty or hostile provider can stream an effectively unbounded JSON d

Article URL: https://www.williamangel.net/blog/2026/06/22/The-AI-Tarpit-Why-You-Cant-Stop-Reading-Your-Code.html Comments URL: https://news.ycombinator.com/item?id=48639685 Points: 1 # Comments: 0

<pre style='white-space:pre-wrap;width:81ex'>fix(agents): bound OpenRouter model catalog response reads (#95420) * fix(agents): bound OpenRouter model catalog response reads The runtime OpenRouter model-capability detector fetched the full /models catalog with an unbounded `await response.json()`, so a compromised or misbehaving endpoint could stream an arbitrarily large body and force the process to buffer the whole payload before parsing. Read the body through the shared bounded reader instead

Article URL: https://www.kcl.ac.uk/news/artificial-intelligence-under-nuclear-pressure-first-large-scale-kings-study-reveals-how-ai-models-reason-and-escalate-under-crisis Comments URL: https://news.ycombinator.com/item?id=48639645 Points: 1 # Comments: 1

<pre style='white-space:pre-wrap;width:81ex'>fix #95489: [Bug]: claude-cli out-of-credits error bypasses model fallback chain — error text delivered as final response (#95508) * fix(agents): fallback on generic cli failure text * fix(agents): guard generic cli failure payload visibility Co-Authored-By: Claude <noreply@anthropic.com> * fix(agents): use exported generic failure text Signed-off-by: sallyom <somalley@redhat.com> --------- Signed-off-by: sallyom <somalley@redhat.com> Co-authored-by:

Article URL: https://github.com/akg268/prompt-preflight/ Comments URL: https://news.ycombinator.com/item?id=48639577 Points: 2 # Comments: 0

Article URL: https://marirs.net.in/cot/ Comments URL: https://news.ycombinator.com/item?id=48639567 Points: 2 # Comments: 1

Used 'potatometer.com' to scan and analyze all All 197 YC Spring 2026 startups on their SEO / GEO / AEO technical setup. I scanned the URL each startup lists in YC's directory. Most are readable by AI crawlers. Most don't tell a crawler what they are. Read more in the blog above. Comments URL: https://news.ycombinator.com/item?id=48639511 Points: 2 # Comments: 0

Article URL: https://aws.amazon.com/about-aws/whats-new/2026/06/aws-lambda-microvms/ Comments URL: https://news.ycombinator.com/item?id=48639498 Points: 19 # Comments: 3