Parameter Efficiency Is Not Memory Efficiency: Rethinking Fine-Tuning for On-Device LLM Adaptation
arXiv:2604.22783v1 Announce Type: new
Abstract: Parameter-Efficient Fine-Tuning (PEFT) has become the standard for adapting large language models (LLMs). In this work we challenge the wide-spread assumption that parameter efficiency equates memory efficiency and on-device adaptability. We show that...
arXiv:2604.22782v1 Announce Type: new
Abstract: Serving transformer language models with high throughput requires caching Key-Values (KVs) to avoid redundant computation during autoregressive generation. The memory footprint of KV caching is significant and heavily impacts serving costs. This work ...
BiTA: Bidirectional Gated Recurrent Unit-Transformer Aggregator in a Temporal Graph Network Framework for Alert Prediction in Computer Networks
arXiv:2604.22781v1 Announce Type: new
Abstract: Proactive alert prediction in computer networks is critical for mitigating evolving cyber threats and enabling timely defensive actions. Temporal Graph Neural Networks (TGNs) provide a principled framework for modeling time-evolving interactions; howe...
KARL: Mitigating Hallucinations in LLMs via Knowledge-Boundary-Aware Reinforcement Learning
arXiv:2604.22779v1 Announce Type: new
Abstract: Enabling large language models (LLMs) to appropriately abstain from answering questions beyond their knowledge is crucial for mitigating hallucinations. While existing reinforcement learning methods foster autonomous abstention, they often compromise ...
The Spectral Lifecycle of Transformer Training: Transient Compression Waves, Persistent Spectral Gradients, and the Q/K--V Asymmetry
arXiv:2604.22778v1 Announce Type: new
Abstract: We present the first systematic study of weight matrix singular value spectra \emph{during} transformer pretraining, tracking full SVD decompositions of every weight matrix at 25-step intervals across three model scales (30M--285M parameters). We disc...
Meet Talkie-1930: A 13B Open-Weight LLM Trained on Pre-1931 English Text for Historical Reasoning and Generalization Research
What if a language model had never heard of the internet, smartphones, or even World War II? That’s not a hypothetical — it’s exactly what a team of researchers led by Nick Levine, David Duvenaud, and Alec Radford has built. They call it talkie, and it may be the most historically disciplined large ...
Build a Reinforcement Learning Powered Agent that Learns to Retrieve Relevant Long-Term Memories for Accurate LLM Question Answering
In this tutorial, we build a Reinforcement Learning–driven agent that learns how to retrieve relevant memories from a long-term memory bank. We start by constructing a synthetic memory dataset and generating queries that require the agent to recall specific information. Using OpenAI embeddings, we c...
OpenMOSS Releases MOSS-Audio: An Open-Source Foundation Model for Speech, Sound, Music, and Time-Aware Audio Reasoning
The model unifies speech, environmental sound, music, and temporal reasoning into a single architecture — and outperforms every open-source model tested on general audio benchmarks, including systems more than four times its size.
The post OpenMOSS Releases MOSS-Audio: An Open-Source Foundation Mode...
DeepMind’s David Silver just raised $1.1B to build an AI that learns without human data
Ineffable Intelligence, a British AI lab founded a mere few months ago by former DeepMind researcher David Silver, has raised $1.1 billion in funding at a valuation of $5.1 billion.
OpenAI is available at FedRAMP Moderate authorization for ChatGPT Enterprise and the OpenAI API, enabling secure AI adoption for U.S. federal agencies.
How Spreadsheets Quietly Cost Supply Chains Millions
A simulation of how a single forecast change moves through five planning teams, and why most retailers lose money in the gap between Sales and Stores.
The post How Spreadsheets Quietly Cost Supply Chains Millions appeared first on Towards Data Science.
Artificial intelligence may be dominating boardroom agendas, but many enterprises are discovering that the biggest obstacle to meaningful adoption is the state of their data. While consumer-facing AI tools have dazzled users with speed and ease, enterprise leaders are discovering that deploying AI a...
Comparing Explicit Measures to Calculation Groups in Tabular Models
With the advent of UDFs and their combination with calculation groups, I see a lot of discussion about not creating explicit measures but instead offering calculation groups to report creators.
The post Comparing Explicit Measures to Calculation Groups in Tabular Models appeared first on Towards Dat...
Show Your Work: The Case for Radical AI Transparency
A colleague told me something recently that I keep thinking about. She said, unprompted, that she appreciated seeing both sides of my AI conversations. Not just the output. The full thread. My prompts, the AI’s responses, the back and forth, the dead ends, the iterations. She said it made her trust ...
Google Deep Research Max: Build Autonomous AI Research Agents in Minutes
Google just changed how developers do research. On April 21, 2026, they launched Deep Research Max. It runs on Gemini 3.1 Pro and is not just another chatbot upgrade. This is an autonomous AI research agent. It plans, searches, reads, reasons, and writes, all from a single API call. By the end, you ...
Meta AI Releases Sapiens2: A High-Resolution Human-Centric Vision Model for Pose, Segmentation, Normals, Pointmap, and Albedo
Meta Reality Labs releases a new foundation model family for human-centric vision that pushes pose estimation, segmentation, and 3D geometry to new state-of-the-art levels — all from a single backbone.
The post Meta AI Releases Sapiens2: A High-Resolution Human-Centric Vision Model for Pose, Segment...