Knowledge catalogue
arXiv:2609.13283v1 Announce Type: cross Abstract: We present a multimodal foundation model for lunar remote sensing, pretrained from scratch on SomBench, a geographically partitioned corpus of nearly
arXiv:2609.01840v1 Announce Type: new Abstract: Spontaneous movement is one of the earliest windows onto an infant's neuromotor health, and structured clinical instruments that score it are validated
arXiv:2608.30963v1 Announce Type: cross Abstract: Modern large language model (LLM) serving systems increasingly operate over repeated or shared context, yet each model typically performs its own pref
Historically, enterprise predictive analytics tasks such as predicting churn, purchase intent, or fraud scoring have meant building custom models using libraries like XGBoost, Random Forest, or Deep N
arXiv:2608.29930v1 Announce Type: new Abstract: Black-box model attribution is increasingly relevant when large language models (LLMs) are served through relay and reseller APIs. A tempting low-cost s
arXiv:2606.16011v2 Announce Type: replace Abstract: Standard accuracy benchmarks evaluate whether large language models (LLMs) reach correct answers. However, they do not test whether models maintain
arXiv:2603.03485v4 Announce Type: replace-cross Abstract: Recent video diffusion models have achieved impressive capabilities as large-scale generative world models. However, these models often strugg
arXiv:2608.26344v1 Announce Type: new Abstract: We previously reported a ModernBERT encoder trained from scratch for Turkish (MoganBERT-TR) and a single-vector embedding model built on top of it (Moga
arXiv:2608.26121v1 Announce Type: new Abstract: Large language models state false facts as fluently as true ones, yet a model often 'knows' internally when it is on shaky ground: the probability it as
NVIDIA TensorRT Model Connect delivers a set of reference implementations for deploying open‑source models from Hugging Face checkpoints to native C++ inference in two steps: a Python CLI builds a dep
arXiv:2608.24168v2 Announce Type: replace Abstract: A unified audio model must recognize and understand linguistic, paralinguistic, and environmental information while supporting speech synthesis and
arXiv:2608.26956v1 Announce Type: new Abstract: Reward models play an essential role in aligning visual generative models, yet most existing visual reward models use a single scalar score or rely on f
arXiv:2608.23641v1 Announce Type: new Abstract: Model welfare research infers what a model prefers from the answers returned to prompts written to elicit preferences. Keeling et al. (2024), Mazeika et
arXiv:2608.24597v1 Announce Type: cross Abstract: Electroencephalography (EEG) is a widely used window into human brain function, but most EEG models remain tied to a one-dataset-one-model supervised
arXiv:2608.14264v1 Announce Type: cross Abstract: Model merging combines trained models directly in weight space, offering a compute-efficient alternative to additional fine-tuning. Selecting merge pa
A model is only as good as its data, and we’ve long since exhausted the internet. From here on out, model progress is gated by data production. @mercor_ai’s @BrendanFoody joined us at our Sovereign AI
arXiv:2608.10107v1 Announce Type: new Abstract: Emerging World-Action Models (WAMs) have demonstrated promising performance in autonomous driving by jointly modeling future driving scene evolution and
Something happened today that I think most people are going to miss because there are two separate stories and neither one is getting the full picture. OpenAI expanded Daybreak. If you haven't heard o
NVIDIA Nemotron 3.5 Lighting is available on Ollama! It's a 30B model made for always-on agents. All local. Claude Code ollama launch claude --model nemotron-3.5-lightning Hermes Agent ollama launch h
arXiv:2608.07040v1 Announce Type: new Abstract: Solving combinatorial optimization problems (COPs) requires not only efficient algorithms but also carefully crafted formulations. While recent works ha
A few months ago, I had the idea of making a LLM from scratch as a personal project (for learning and partly for improving my resume). Since I learned a lot from other posts on here over the past year
What Local Embedding + Reranking Models are you guys running for RAG? I went down this rabbit hole because I wanted a Embedding Model + Reranker for a Translation Memory Server. Essentially, given X p
arXiv:2608.04215v1 Announce Type: cross Abstract: The growing diversity of code clone types, from syntactic copies to cross-language semantic clones to AI-generated duplicates, has created a fragmenta
We developed a leaderboard ranking frontier model security. There's no shortage of model capability rankings, but we didn't find anything comparable for model security. Yet security is becoming increa
arXiv:2607.26023v1 Announce Type: new Abstract: Graph foundation models (GFMs) have emerged as a promising paradigm for transferring knowledge across graph domains and tasks. Real-world graphs associa
Since I am needling every model maker tonight about minor but important issues, one more: Grok 4.5 has no model card. Companies that are trying to compete near the frontier should be releasing model c
arXiv:2603.14161v2 Announce Type: replace Abstract: Many disciplines need quantitative models that synthesize experimental data across multiple instances of the same general system. For example, neuro
arXiv:2510.06048v4 Announce Type: replace Abstract: Effective data selection is essential for pretraining large language models (LLMs), enhancing efficiency and improving generalization to downstream
Pretty refreshing for foundation model training in 2026. I wonder how much of the capability gap between their model and other frontier models is because of this constraint. from: https://microsoft.ai
arXiv:2602.03639v2 Announce Type: replace Abstract: Sampling-based controllers, such as Model Predictive Path Integral (MPPI) methods, offer substantial flexibility but often suffer from high variance
arXiv:2602.11137v2 Announce Type: replace-cross Abstract: Large language models are typically trained in two broad phases: pretraining to produce a base model, followed by further training to improve
We are releasing Carbon: a crazy fast DNA model Carbon is 275x faster than the next best model. So fast you can process the whole human genome on a single GPU in <2 days. Here are the tricks we used:
arXiv:2605.10447v1 Announce Type: cross Abstract: Agent-based models (ABMs) are increasingly used in macroeconomics, but their analysis still often relies on ad hoc Monte Carlo campaigns with heteroge
the deeper point: model choice and harness choice aren't independent variables. benchmarking a model without specifying the harness is like benchmarking a chip without specifying the compiler. the num
arXiv:2605.00994v1 Announce Type: new Abstract: Finetuning can significantly modify the behavior of large language models, including introducing harmful or unsafe behaviors. To study these risks, rese
The year of the harness 🤝 Harnesses are our most direct layer to turn a model into a great product experience for users by shaping model behavior towards a goal This matters today because models have
Vincent Chow / South China Morning Post: Tencent releases Hy3-preview, its first AI model developed under former OpenAI researcher Yao Shunyu; the model features 295B parameters, down from HY2's 400B
I tested @huggingface ml-intern, given the prompt 'Fine-tune a Segment Anything Model (SAM) on a useful medical dataset. Train the model, and provide a comprehensive tutorial in a Jupyter Notebook fil
Auto-generated index of all papers mentioned across the wiki.
Anthropic investigated the internal mechanisms of its latest unreleased model, Claude Mythos Preview, and what they found is 100% worth a read. Key things I pulled from Anthropic researchers' threads:
arXiv:2610.14937v1 Announce Type: new Abstract: How do language models generalize from their training data? This is a central challenge for alignment: models reliably learn their training distribution
arXiv:2610.13134v1 Announce Type: new Abstract: We present Neurograf v1, an experimental end-to-end Romanian-first language-model pipeline developed to investigate the feasibility of training language
arXiv:2610.07286v1 Announce Type: cross Abstract: Model optimizations help improve inference performance and accuracy of ML workflows. However, relying on a single model to perform inference across al
arXiv:2603.23184v2 Announce Type: replace-cross Abstract: Despite the success of reinforcement learning from human feedback (RLHF), existing reward modeling methods largely rely on explicit feedback,
arXiv:2610.05095v1 Announce Type: cross Abstract: Latent Dirichlet Allocation (LDA) and models derived from it remain widely used topic models. LDA observes each document as a bag-of-words and models
arXiv:2610.02439v1 Announce Type: new Abstract: Generative graph models are central to understanding and simulating complex networks. However, existing approaches have complementary strengths and limi
Hi r/LocalLLaMA. I'm on the team at Blockway, a small team in Hong Kong (disclosure: this is our model). Today we released Agens Volundr 32B Preview, the first model built on our own hybrid architectu
Somebody just released a fully open-source, open-weights decision model that beats Jev! Just kidding, it's me and I this is my first time releasing a public model, recipe and dataset so I am looking f
I've been using my free time during weeknights and weekends for the last 11 months working on and refining a small domain-specific language model. It specializes on information about World War II. The
Hi again! This is a update to my previous post MageTrail V0.25 where I continued this tech demo toy model thingy called MageTrail, a Danbooru/E621 proof of concept Full-Finetune of Microsoft's MageFlo
Good day data nerds! I am trying to get more into running models, learning about agentic workflows, and creating my own tools. But as you do (right?) I had to fine tune my current setup. With the way
Since my last post, I've been thinking about different options for dynamic performance degradation, trying to squeeze as much high-quality inference out of my GPU as I can. Over the weekend I read thi
test-save-load-state : print a per-model results table in --models mode (#29316) test-save-load-state : print a per-model results table in --models mode in --models mode the output was very heavy: eve
arXiv:2505.10083v2 Announce Type: replace Abstract: Conventional forecasting methods are trained end-to-end on unimodal time series, which limits their ability to exploit textual information and under
Introducing Contrastive Language Model (CLM): an ultra-fast System One Model trained with a contrastive learning objective that connects states and actions. CLM-8B is pre-trained on internet-scale dat
Overview llama.cpp 0.4.1 adds Maple 20B-A1B, Tencent Hy 4, and Spark2.5 support, improves JSON schema handling, chat parsing, logging, and server child-process management, and updates ggml to v0.24.0.
I have a full model, it's ready to train. It's ~9b parameters. 9.4b to be more exact. That includes a 1/2/3 Engram table, Moonshot's AttnRes modeling, and RoPE / NoPE layering at 3:1 as more or less v
arXiv:2609.11446v1 Announce Type: new Abstract: Heterogeneous model collaboration seeks to exploit the complementary strengths of different models to balance predictive performance and inference cost.
model: add Kimi-K3 text model (#26185) model: add Kimi-K3 text model Hybrid KDA (linear) + MLA (full) attention as in Kimi-Linear-48B, plus five things that architecture does not have: cross-layer res
arXiv:2608.13063v1 Announce Type: new Abstract: Prior work on LLM behavior under anomalous conditions asks whether a model notices anomalies. We ask a narrower question: once a model sits in a workflo