All entries

The problem we face with AI today is not a technical problem, it is a political problem. Of who gets to decide. What level of risk the public is exposed to, what future we build towards or avert. It's

arXiv:2606.23897v1 Announce Type: cross Abstract: Prompt distillation compresses large vision-language models (VLMs) such as CLIP into lightweight student models by matching teacher predictions on unl

arXiv:2504.17768v3 Announce Type: replace Abstract: Sparse attention offers a promising strategy to extend long-context capabilities in Transformer LLMs, yet its efficiency-accuracy trade-offs remain

The strange thing about progress is that it only counts if you ignore where we started. Every civilisation practised slavery. The civilisation that ended it is now told it should feel guilty forever f

Samantha Kelly / Bloomberg: The US FDA drops an enforcement complaint against Whoop over its blood pressure tracking tool, reversing a July 2025 warning letter; Whoop is updating the tool — The US Foo

The users who complain about the flaws in your product may seem annoying, but they are on the whole probably your most valuable users. They complain because they care, and I doubt a startup could ever

Clem Delangue discusses a resurgence of interest in decentralized intelligence systems and approaches, suggesting a shift away from centralized AI models toward more distributed architectures. The pos

arXiv:2606.24627v1 Announce Type: new Abstract: Fact-checking systems built on LLMs achieve high verdict accuracy on standard benchmarks, yet routinely output Supports labels whose cited evidence does

Gary Marcus argues that generative AI, while significant, represents only a narrow subset of the broader computing field and should not overshadow the diverse and impactful work occurring across other

arXiv:2606.24622v1 Announce Type: new Abstract: Training safe Reinforcement Learning (RL) systems is inherently challenging, with no guarantee of avoiding unwanted behaviors. The most effective defens

Thinking about AI Engineer World's Fair? Apply to be an Associate! It's a great way to connect with fascinating people across the AI ecosystem, build new relationships, and be part of one of the most

arXiv:2511.07397v2 Announce Type: replace Abstract: Voice agents face a fundamental tension: the reasoning, retrieval, and tool use that make foundation models capable are iterative and slow, while co

As soon as August, a giant silver bullet will cut its way through the dry air of the southwestern US and cross the Pacific to reach the coast of Japan. Once there, the roughly 200-foot-long craft, bui

This is one of those moments where I feel most disconnected from my normie friends I'm like 'do you realize what this means?' And they say 'no' and I realize I will need to explain decades of nerd lor

This is the strongest ARC-AGI-2 performance to date by an open-source model. GLM-5.2 from @Zai_org on ARC-AGI (Verified) - ARC-AGI-2: 22.8%, 0.25 - ARC-AGI-1: 77.0%, 0.19 Performance is comparable wit

arXiv:2606.24336v1 Announce Type: new Abstract: Face Video Restoration (FVR) aims to recover high-fidelity facial videos from degraded input while preserving identity and semantic consistency across f

Yohei Nakajima discussed Activegraph at an AI House event, likely covering applications or insights related to graph-based AI systems or knowledge representation. This appears to be a social media pos

arXiv:2606.18729v2 Announce Type: replace-cross Abstract: Data valuation quantifies the intrinsic quality of individual samples to enable principled data curation, quality control, and robust learning

arXiv:2606.24596v1 Announce Type: new Abstract: As Large Language Models are increasingly deployed in critical applications, robustly evaluating their social biases is paramount. However, the current

Nous Research recommends using the 'hermes skills config' command to enable creative-ideation and other optional skills in their system. This command appears to be a configuration tool for activating

Today I present you a curated collection of @tszzl saying that Fukuyama is always right We Need an International Treaty to Ban Superintelligence https://open.substack.com/pub/persuasion1/p/time-for-a-

arXiv:2606.24074v1 Announce Type: cross Abstract: Wang~ite{Wang2026} introduced the Stochastic-Oracle Turing Machine (SOTM) framework and defined token complexity as the minimum expected cost of inter

arXiv:2606.24021v1 Announce Type: new Abstract: In modern generative models, images are specified and controlled through text prompts. In practice, images are generated from sequences of tokens derive

arXiv:2606.23901v1 Announce Type: new Abstract: This paper addresses the problem of robust formation control by introducing Topological Online Learning for Displacement-based (TOLD) formation control,

arXiv:2606.24598v1 Announce Type: cross Abstract: While expert-validated 'LLM + script' workflows deliver significant value, they remain static: they encode hard-won domain knowledge yet fail to adapt

arXiv:2606.24187v1 Announce Type: new Abstract: Long video understanding remains a daunting challenge for Multimodal Large Language Models (MLLMs) due to the excessive computation and memory footprint

arXiv:2606.24237v1 Announce Type: new Abstract: Federated Graph Learning facilitates collaborative graph modeling across distributed clients while preserving data privacy. However, real-world data cat

arXiv:2606.24004v1 Announce Type: cross Abstract: Steering a large language model (LLM) toward a desired behavior typically relies on an iterative process of hand-crafting a prompt based on a careful

arXiv:2606.24040v1 Announce Type: cross Abstract: MeMo proposes language models with explicit multi-layer correlation matrix memories (CMMs), where memorization, retrieval, and forgetting are architec

Tracing functionality can be utilized to monitor and analyze memory usage patterns in applications, helping developers identify memory leaks, optimize allocation, and understand runtime behavior. This

arXiv:2606.24279v1 Announce Type: new Abstract: In Description Logics (DLs), reasoning under Rational Closure (RC) is a well-known and widely accepted non-monotonic formalism to handle defeasible know

arXiv:2606.24297v1 Announce Type: new Abstract: Cross-domain Few-shot Segmentation (CD-FSS) aims to transfer knowledge learned from source domain to distinct target domains, segmenting unseen target c

Traitor-maxxing August, 2021. Mayorkas goes to the Southern border and announces a bunch of USAID funded programs to improve the lives of the people in Guatemala, Honduras, and El Salvador. We later f

arXiv:2606.24430v1 Announce Type: cross Abstract: Training of neural networks for histopathology classification tasks typically relies on data encoding into latent space, which reduces complexity and

arXiv:2606.24331v1 Announce Type: new Abstract: Transformer-based language models have become the default substrate for natural language processing and the pace of new releases has made it hard for pr

Brock E.W. Turner / Axios: Trase, which is building an operating system and infrastructure layer for AI agents in industries like health care and defense, raised a 107M seed led by Arch — Trase, an AI

arXiv:2606.24175v1 Announce Type: new Abstract: While point cloud foundation models have significantly advanced point cloud video understanding, existing parameter-efficient fine-tuning (PEFT) methods

arXiv:2606.24256v1 Announce Type: new Abstract: Physical interactions follow a long-tailed distribution: a set of common and regular interactions dominates human experience and visual data, while a br

arXiv:2606.24302v1 Announce Type: new Abstract: Fine-tuning transformer-based handwritten text recognition (HTR) models on medieval manuscripts is challenging because these models are pre-trained on m

arXiv:2605.13076v2 Announce Type: replace Abstract: The LLM-based generation of machine-readable outputs such as JSON has attracted significant attention for integration with external systems. However

arXiv:2606.23917v1 Announce Type: new Abstract: Advances in generative AI have made image falsification highly realistic, demanding trustworthy authentication systems. Existing forensic detectors can

Try Grok I switched the Grok as my search engine a few months ago. I only use Google now to correct typos / check spelling, as it's slower to use Grok for those tasks. Grok is superior to Google in ev

Try Kimi K2.7 and GLM 5.2 for free in Devin Desktop and CLI Kimi K2.7 Code and GLM 5.2 are available in Devin Desktop and CLI Both perform strongly on FrontierCode Extended, our benchmark for real-wor

arXiv:2507.01752v4 Announce Type: replace-cross Abstract: Gradient-based optimization is the workhorse of deep learning, offering efficient and scalable training via backpropagation. However, exposing

arXiv:2606.24039v1 Announce Type: new Abstract: Robotics increasingly relies on GPUs for parallel simulation, large-scale learning, and neural-network inference. For model predictive control (MPC) to

arXiv:2606.24253v1 Announce Type: new Abstract: Modern VLMs and VLA systems commonly adopt off-the-shelf ViTs such as SigLIP2 as visual encoders, but diverse downstream requirements in latency, tempor