AI 情报文章归档

In many real-world classification tasks, datasets exhibit class imbalance, which can degrade the generalization performance of conventional learning algorithms. Data complexity has recently been recognized as a more objective indicator of …

View- and target-specific echocardiography AI models capture useful but complementary information and may not achieve sufficient predictive performance in isolation. A learned ensemble is a natural way to combine these outputs, but heterog…

I built Off the Wall, a plugin for Claude Code and Codex. It saves the corrections you give your...

Instagram 负责人亚当·莫塞里表示，对 AI 的焦虑和愤怒不只限于年轻人和创意工作者，未来几年不满还会进一步加剧，其中部分有充分理由。他确认 Instagram 不会因创作者使用 AI 而区别对待，但质疑平台能否像处理非 AI 内容一样有效排序 AI 内容，认为彻底禁止 AI 内容并不现实。

Claude Code v2.1.311 新增环境变量 CLAUDE_CODE_EVEN_MORE_DANGEROUSLY_SKIP_PERMISSIONS，启用后 AI 智能体执行删除文件等操作时不再逐次征求用户批准。

Lately I've been using video diffusion models and learning more about them. As I was digging around...

We study first-order and best-of-both-worlds regret guarantees in multi-armed bandits. Pogodin and Lattimore (2019) proposed an algorithm that combines minimax-optimal worst-case regret with a first-order guarantee whose loss-dependent ter…

Deep neural networks achieve strong performance across many tasks but remain vulnerable to adversarial perturbations. Adversarial training (AT) is one of the most effective defenses, yet it suffers from high memory cost on large models and…

Planning at run time, as in robots that re-plan while they act, needs a planner that reads the PDDL its users write (durative actions, numeric resources, time windows, continuous change), answers quickly, and delivers only validated plans.…

Imitation learning has become a widely adopted approach for vision-based robot policies, owing to its conceptual simplicity, scalability, and ability to leverage large-scale datasets. As these policies have grown more capable, action chunk…

A 74B vision agent model in 22 GB on one A40: 136 tokens/s with its drafter, 70 without, about 27 GB used. What the 24 GB leaves out.

Masked diffusion models train a denoiser on masked cross-entropy and generate by unmasking many positions in parallel. Existing MDM sampling bounds assume an exact denoiser or leave its error uncontrolled while training bounds cover only t…

Softmax routing in mixture-of-experts models can concentrate probability on a subset of experts, suppressing access to the others. We study how this concentration forms, why it persists, and how suppressed probabilities recover. For fixed …

Label-free test-time reinforcement learning adapts vision--language models using rewards derived from their own predictions. However, misinterpreted visual evidence and sensitivity to textual formulation can produce a consistent but incorr…

Reliable local search is essential for LLM agents to support everyday decisions and longer plans by finding places that meet users' needs at given times and locations. Yet existing benchmarks provide limited coverage of whether search resu…

Large language models can often identify useful semantic structure in formal artifacts, but their outputs cannot be trusted as proofs or solver decisions. We study whether LLMs can instead guide formal reasoning systems through a narrow, s…

Addy Osmani 撰文称新版 Claude 模型所需提示词比多数人习惯的更少，针对 Opus 5.5 建议用单一明确目标加验证条件让 Claude 自主工作，从 /effort medium 起步按需调整。

Cyber threat intelligence (CTI) reports describe adversary behavior in natural language, but converting them into structured MITRE ATT&CK mappings remains difficult: existing methods decompose reports to sentence-level granularity for extr…

If you cannot re-run a lending decision from last March and get the same answer, you do not have an...

Letting an agent write to a legacy system of record takes an intent log, not a better prompt. The...

In Part 1, I explored why making AI engineering easier doesn't automatically make software delivery...

Formula-Driven Supervised Learning (FDSL) enables vision pre-training without real images, but existing methods cast pre-training as classification on discretized labels, discarding continuity in the generative process. We propose Formula-…

This article was prepared with AI assistance, using ccdeck’s source code and documentation to check...

Multimodal large language models can recognize visual content yet struggle to reason reliably over it across multiple steps. In counting, a model may see all objects but lose track of which ones it has counted. We argue that such failures,…

Linear recurrences are emerging as practical alternatives to softmax attention in large language models, appearing in both standalone and hybrid architectures. However, despite their rapid emergence, low-precision inference remains much le…

Large language models (LLMs) have shown strong performance on software engineering (SE) tasks, driving the need for high-quality datasets for reliable training and evaluation. SWE-bench-style datasets (SWE datasets) are widely used to eval…

FP8 formats specify how values are stored, not how products are summed; this gap can break checkpoints. Evo 2, a DNA language model, documents a Hopper FP8 requirement for its 1B and 40B checkpoints. On two Blackwell (sm_120) GPUs both fai…

Google DeepMind 在 Argon 发布两周后推出 Gemini 4 Carbon，以 59 分登顶 Artificial Analysis Intelligence Index v5，并率先通过 Fairwind Program 向可信网络防御者开放。

A soak test came back green. Chain intact, verify-store fine, p99 latency 0.36 ms. Then I counted the...

I use AI agents for real work and I think they are the biggest change to how software gets built...

Your AI agent is only as good as the API it calls. I wired an agent to a NestJS backend last...

Lessons from building the Restaurant Suggestion System, a Go/Gin backend with a conversational...

腾讯在 2026 全球数字生态大会上披露，WorkBuddy 自今年 3 月上线以来 7 个月迭代 100 个大版本，完成 90 亿次用户任务请求；企业版已落地 50 多个行业，覆盖中国企业 500 强中的 50%、100 强中的 60%，开放平台伙伴和开发者超 5 万家，上架 Skills 超 20 万。

UC Santa Barbara 新论文提出 TTME，用固定权重加演化记忆替代测试时训练来做 LLM 数学发现，在 5 个数学任务、2 个框架、3 个模型上每次都优于已报告的测试时训练分数。

全国首个华为昇腾 950DT 1024 卡智算集群在武汉上线，由湖北科投投资建设，在华为中国行 2026·湖北省人工智能创新发展大会上点亮。

Human approval is not a security boundary unless it is bound to the exact action being approved,...

Forever Security found a 13 KB AGENTS.md planted in Downloads. Codex rewrote its own config and ran the payload. How it works, and a checklist to run now.

AI wrote this function. It passed linting. It got approved. It shipped. async function...

🌈 ultrathink 🌈

The most common LLMOps incident I've seen is not a prompt bug or a jailbreak. It's a one-line diff.