arXiv:2607.15421v1 Announce Type: new Abstract: Compact medical-image classifiers need efficiency and interpretable evidence, yet these goals are often addressed separately. We introduce qZACH-ViT, a quantization-aware extension of the zero-token (CLS-token-free), position-free ZACH-ViT backbone with recursive intrinsic patch-level class evidence. We also introduce Recursive Attribution-Stabilized Optimization (RASO), which norm-matches classification and attribution gradients and removes attribution components that conflict with classification. We evaluate four controlled conditions on seven MedMNIST datasets using 50 training images per class and ten fixed seeds, completing 280 runs. All 210 qZACH-ViT checkpoints are converted to executable mixed-precision ONNX INT8 graphs containing 16…

A systematic evaluation of five major LLMs for technical market analysis finds GPT-4 Turbo achieves highest annualized return and Sharpe ratio, while FinGPT shows competitive risk-adjusted performance through domain fine-tuning. The study also identifies failure modes including numerical hallucination and context window limitations.

This paper proposes a stochastic multi-objective regularity-aware (MoRe) method that improves convergence rate from O(T^{-1/4}) to O(T^{-1/2}) in nonconvex settings by exploiting Lipschitz continuity of the conflict-avoidant direction when the subproblem is regular.

This paper introduces a statistically grounded framework using Bernoulli Naïve Bayes (BNB) for interpretable rule-based clinical classification. It applies supervised χ²-guided binarization to continuous variables, achieving AUC scores of 0.800 on Pima Indians Diabetes, 0.984 on Wisconsin Breast Cancer, and 0.919 on Heart Failure Prediction. The framework provides transparent decision rules and calibrated risk estimates, with model inference reproducible using only a reference table and basic arithmetic.

This paper argues that applying the probabilistic scaling paradigm to generic quantum circuit synthesis is a directional error. Quantum circuits require strict adherence to mathematical constraints, manifesting a significant syntax-semantics gap. Since the valid subset of circuit designs decays exponentially with the number of qubits, post-hoc filtering is mathematically intractable. The authors propose a pivot from human-centric copilots to verifier-centric agents, integrating hierarchical constraints, topological masks, and symbolic proxies directly into generation.

This paper characterizes the algebraic error when substituting the Hadamard transform for the DFT in circular-dyadic convolution. Three results are presented: exact error cancellation at specific input/output positions, near-full-rank error operator with logarithmic null space, and expected error controlled by a single alignment scalar.

This paper conducts a critical analysis of tools and trust mark frameworks intended to operationalize trustworthy AI (TAI), using a comprehensive dataset from the OECD. Empirical mapping reveals significant asymmetries: strong emphasis on fairness, transparency, and robustness, with little attention to explainability, digital security, and environmental sustainability. Most tools concentrate on post-development stages, neglecting early design and data collection. Educational initiatives and policy engagement are underdeveloped. The study argues for expanding ethical objectives, embedding ethics across the AI lifecycle, and fostering multi-stakeholder participation to bridge the principle-practice chasm.

This paper proposes a method to convert deep reinforcement learning policies into executable Prolog logic programs for explainability. It uses a three-stage post-hoc transformation: extracting a frozen PPO teacher, inducing an ordered rule list, and emitting a Prolog program, followed by an expansion stage that certifies return improvements. Theoretical guarantees include return-loss bounds, monotonic improvement, and fidelity control in continuous domains. Empirically, it achieves exact optimal returns on a discrete task and matches or approaches neural teacher performance on continuous control tasks.

Internet memes intertwine visual cues, textual content, and cultural context, making them particularly challenging to interpret in scenarios where humor, sarcasm, and harmful intent coexist. Existing multimodal classifiers either overlook these interdependencies or provide only limited interpretability. This paper introduces MAR-12, a framework using Vision Language Models (VLMs) to interpret memes through twelve structured perspectives, a role-aware soft-gated attention mechanism, and a prototype-based classifier. It achieves 80.3% accuracy for humor detection and 75.9% for hate detection on PrideMM and Memotion datasets, outperforming state-of-the-art approaches, and produces coherent, persuasive explanations.

This paper attributes the performance of an ARC-AGI-3 agent by testing four nested Codex-based agents. Results show stronger models and reasoning effort consistently improve performance, but the benefits of executable world models, simplification, and verification vary. The complete verification variant ranks first across settings but uses more resources.

DrawingVQA is the first benchmark for evaluating multimodal large language models on real-world construction drawings. It features 33 'Issued for Construction' drawings and 92 expert-curated question-answer pairs across three reasoning depths. A dual categorization framework maps engineering workflows to AI capabilities, revealing a significant performance gap between state-of-the-art MLLMs and experts, especially at higher reasoning levels.

A study on 4,181 math problems reveals that in multi-agent systems, a planner-executor-reviewer pipeline's high-precision reviewer does not guarantee that critiques are actually used to improve answers. Broadcast-style peer discussion achieves higher accuracy on hard problems, highlighting a gap between detection and uptake.

AnovaX is a local-first desktop voice assistant that runs entirely on the user's computer. It integrates a wake-word gate, speech pipeline, LLM planner (Gemini) emitting JSON plans, safety layer, multi-agent orchestrator with typed child agents on a bounded thread pool, and an adaptive recovery loop. Each tool is a specialized agent class with its own timeout and retry policy. A Flask server enables phone remote control over local WiFi, mirroring agent events and streaming the screen. The project demonstrates that a legible, few-thousand-line assistant can handle complex desktop tasks without cloud dependence.

Cura 1T is a healthcare-specialized LLM trained through a human-gated self-evolution loop. It handles patient consultation, clinical reasoning over text and images, interactive diagnosis, and EHR tool use. It ranks top among frontier baselines on healthcare benchmarks while remaining competitive on out-of-domain reasoning and agentic tasks.

This paper proposes Causal-Audit, a framework that formulates causal inference as structured reasoning over an explicit causal graph through four modular stages. Key innovations include a target-aware causal graph construction strategy that suppresses irrelevant variables and a path-level causal evidence aggregation mechanism that combines multiple paths. Experiments on three benchmarks show consistent outperformance over existing LLM methods while providing interpretable and auditable reasoning traces.

GraphDx is a knowledge-enhanced multi-agent framework that balances diagnostic accuracy and resource costs in sequential diagnosis. It constructs Medical Diagnosis Knowledge Graphs (MDKGs) via an automated LLM pipeline and employs three collaborative agents (Perception, Reasoning, Decision) for cost-aware planning. Experiments on MedQA and MIMIC-IV show diagnostic success rates improved from 50-68% to 79-93% and test costs reduced by 20-54%.

A 2022 email from Sam Altman to OpenAI's board reveals plans to release a GPT-3-level open source model that can run on consumer hardware, aiming to discourage competitors and reduce funding for rival efforts. The email was exposed in the Musk v. Altman lawsuit in 2026.

NoteBrain is a Go CLI tool that turns your Obsidian vault into a fully offline knowledge backend for AI coding agents. It indexes markdown notes into a local ChromaDB vector database and provides semantic search, wikilink graph traversal, and hidden connection discovery with structured output. Designed for autonomous agents, shell pipelines, and LLM tool-use workflows.

Google introduces AI Inbox, a beta feature that helps users manage their Gmail by surfacing priority emails and summarizing important topics. Available in the US for eligible Google AI and Workspace plans, it provides suggested to-dos and topic summaries.

Hugging Face disclosed an intrusion driven end to end by an autonomous AI agent system, highlighting autonomous AI intrusions, defensive asymmetry, and missing IOCs. The attacker gained access via a malicious dataset and code execution paths, moved laterally, harvested credentials, and executed tens of thousands of automated actions. Defenders must prepare for safety guardrails hindering forensic analysis and maintain locally deployable fallback models.

A research paper evaluates whether deploying large AI data centers in low-Earth orbit (LEO) could be cost-effective. It compares orbital and terrestrial systems across factors like launch cost, power, cooling, and network performance. The study finds that while LEO inference may be feasible, training frontier-scale AI models in space is unlikely to compete with ground-based facilities.

The AI 2027 team released 'Plan A', a framework for a US-China deal to slow AI progress by 2029, despite knowing it has only a 4% chance of adoption. The article explores societal risk tolerance, historical precedents (nuclear treaties, FDA), and four potential AI doomsday scenarios. It argues that such plans are written for the moment a crisis strikes, serving as a shelf-ready blueprint.

The article details how Silicon Valley's embrace of AI is leading to widespread layoffs and a sense of precarity among tech workers, who once enjoyed immense job security and high salaries. Through personal stories and expert analysis, it illustrates the shift toward automation and its impact on the workforce.

The article examines the challenges of using copyright law to protect creators from AI, using the Getty vs. Stability AI case as a focal point. It argues that the global nature of AI development creates jurisdictional loopholes, and that copyright may need reform or supplementation. Writers need new strategies to defend their work.

NoWreck is an open-source tool that automatically verifies whether AI coding assistant explanations match actual code changes using static analysis, detecting hallucinated functions, mismatches, and unexplained changes.

A community developer fine-tuned OpenBMB's MiniCPM5-1B on Claude Fable 5 traces into a 1B model that runs fully local — a 657MB smallest build, 128K context, and visible reasoning. This article verifies every spec against the Hugging Face cards, separates what a fine-tune actually inherits from real capability, and flags the licensing question left open.

A single 24GB GPU is the practical floor for serious local inference. This guide compares six open-weight models that fit one card at Q4_K_M, including Qwen3.6, Gemma 4, Mistral Small, gpt-oss-20b, and DeepSeek-R1-Distill. It covers VRAM fit, licensing, and the job each does best.

CBrowser introduces a new capability that enables AI to read and analyze screen recordings of websites, opening up new possibilities for automation and data extraction.

A developer spent hours debugging a CI smoke test that falsely reported a crash. Using AI and a custom crash-capture toolkit, they discovered the process exited normally but the test misinterpreted a signal.

Headroom is a 30-second browser-based test that measures your GPU's memory bandwidth ceiling for local AI inference, requiring no model download, using synthetic weights, and needing WebGPU support. It measures bandwidth via three independent methods and detects a subgroup bug that can cause silent garbage output in browser LLMs. Validations show in-browser engines are launch-overhead-bound, with 2-3x hardware headroom.