Knowledge catalogue
arXiv:2610.17726v1 Announce Type: new Abstract: Large language model (LLM) agents excel at complex, multi-step tasks but suffer from catastrophic forgetting during continual learning--acquiring new sk
arXiv:2610.15093v1 Announce Type: cross Abstract: AI coding agents operate in iterative loops: they modify code, run tools, inspect the results, and decide what to do next. We call the design of this
arXiv:2610.18516v1 Announce Type: new Abstract: Post-training pruning enables efficient deployment, but conventional methods do not explicitly preserve internal-gradient alignment between sparse and d
arXiv:2603.18572v2 Announce Type: replace-cross Abstract: Deep unrolled models (DUM) have become the state-of-the-art for accelerated MRI reconstruction, yet their robustness under domain shift remain
arXiv:2610.14696v1 Announce Type: new Abstract: Bayesian optimization uses posterior uncertainty to decide where to explore, but this uncertainty is not spatially neutral. On bounded domains, Gaussian
arXiv:2610.12941v2 Announce Type: replace Abstract: Image deblurring is a fundamental task in computer vision, yet existing methods mainly specialize in isolated scenarios. Consequently, the deblurrin
arXiv:2610.16969v1 Announce Type: cross Abstract: Latent Action Models (LAMs) can capture manipulation priors from action-free videos, but discrete latents usually impose a granularity bottleneck, lim
arXiv:2610.16707v1 Announce Type: cross Abstract: This paper considers a distributed estimation system where edge nodes transmit averaged features to a center node for parameter estimation. A key chal
arXiv:2204.09334v4 Announce Type: replace-cross Abstract: Unsupervised domain adaptation approaches have recently succeeded in various medical image segmentation tasks. The reported works often tackle
[UPDATE / RETRACTION: The '100% GTO' claim had a prompt-leakage bug] We patched the repo to pass strictly neutral states, here are the real, honest numbers: • Latency: Still ~2ms (ONNX + Absorbing Mar
arXiv:2610.17252v1 Announce Type: cross Abstract: Search-based policy learning, such as Monte Carlo tree search (MCTS), offers the flexibility to trade runtime computation for data efficiency and adap
arXiv:2610.17383v1 Announce Type: new Abstract: Large-scale repositories of wearable sensor data have enabled the development of foundation models for human activities and behavior. However, existing
Very cool to see Embedding Gemma 2 by @Google, number on trending on HF! The media and silicon valley bubble only talks about reasoning LLMs when there is a much bigger variety of AI models that are s
arXiv:2610.14158v1 Announce Type: new Abstract: Understanding and tracking a dynamic world are fundamental to visual intelligence. Existing video benchmarks provide limited insight into whether models
arXiv:2610.15389v1 Announce Type: new Abstract: Video understanding agents rely on a harness to acquire evidence, use tools, and reason, but improving it requires substantial effort. Recent self evolu
arXiv:2610.15915v1 Announce Type: new Abstract: Reusable skills improve multimodal agents without updating model weights, but their effectiveness depends on both the evidence presented and the procedu
arXiv:2610.16441v1 Announce Type: new Abstract: Current image-text evaluation often relies on a global score, which gives limited insight into which semantic details are supported by the image. Explai
arXiv:2610.15556v1 Announce Type: cross Abstract: Vision-Language-Action (VLA) models commonly generate robot actions token by token through Large Language Model (LLM) decoders, causing inference late
arXiv:2610.14403v1 Announce Type: new Abstract: Learning reusable skills can improve visual reasoning agents, yet applying a skill still requires the agent to implement or invoke a procedure and produ
arXiv:2610.14731v1 Announce Type: new Abstract: Long-term memory benchmarks for language agents primarily evaluate whether information from earlier interactions can be retrieved and used when relevant
arXiv:2610.17057v1 Announce Type: cross Abstract: Vision-based document retrieval has improved rapidly, but current evaluation settings still emphasize short documents and single relevant pages. This
arXiv:2602.05382v2 Announce Type: replace Abstract: Vision-language model (VLM) errors on visual reasoning may come from missing visual structure, from mistakes after that structure is extracted, or f
arXiv:2610.17986v1 Announce Type: new Abstract: A common practice in instance retrieval is to grow the gallery size to improve a benchmark's discrimination capability. One prominent example is ILIAS,
We just shipped web search in Private Mind, our open source app for running LLMs locally on iOS and Android. The video is the whole pitch: Qwen 3 1.7B with web off makes up Apple's September event, wi
Welcome to Mistral, New Material. Let’s bring design to mission-critical AI. New Material joins @MistralAI to design the frontier. Over the last 6 years we've teamed up with ambitious organisations tr
arXiv:2610.15283v1 Announce Type: cross Abstract: In off-policy evaluation, merging decision-time histories that matter differently for a specified policy value or comparison can make that evaluation
What AI Safety needs is better tooling and engineering. For instance, building stronger red-teaming tools. Great to see this open-source effort to red-team image models. It seems to learn better ways
arXiv:2610.14007v1 Announce Type: new Abstract: A familiar line of reasoning runs as follows: language models are flexible approximators trained under a strictly proper loss, so scale brings them clos
arXiv:2610.15033v1 Announce Type: cross Abstract: Generative search engines such as ChatGPT, Gemini and Perplexity now decide which Web pages users see as sources, and a growing industry of generative
arXiv:2610.18869v1 Announce Type: cross Abstract: How should demonstration quality be used during vision-language-action (VLA) pretraining to improve performance after target-task fine-tuning? We comp
arXiv:2610.15778v1 Announce Type: new Abstract: When a policy is trained with GRPO against a frozen, reference-free LLM judge, the judge accepts more of its answers, but in a post-hoc analysis we dete
arXiv:2610.14530v1 Announce Type: new Abstract: Multi-robot systems (MRS) offer significant advantages over single-robot platforms by enabling information sharing and coordination for different tasks
arXiv:2610.16152v1 Announce Type: new Abstract: Although post-training significantly enhances model performance, the use of massive undisclosed data raises concerns about benchmark contamination and c
arXiv:2610.17355v1 Announce Type: new Abstract: Video language models represent a video as a long sequence of visual embeddings in a decoder's input. Every visual embedding is processed by every decod
arXiv:2606.18097v2 Announce Type: replace Abstract: Deformable Linear Objects (DLOs), such as wires and cables, are central to industrial assembly. Unlike rigid objects, whose state is captured by a 6
arXiv:2610.17434v1 Announce Type: new Abstract: Recent advances in video world models have enabled plausible interactive environments, yet long-horizon generation places a fundamental demand on visual
arXiv:2610.18826v1 Announce Type: cross Abstract: We introduce WorldOracle, a benchmark specifically designed to evaluate action-conditioned world models (ACWMs) against paired real-world executions.
arXiv:2610.15394v1 Announce Type: new Abstract: Structured pruning is typically formulated as identifying which model dimensions to remove, often within the model's original representation. However, t
arXiv:2610.16131v1 Announce Type: new Abstract: Financial forecasting has traditionally relied on numerical time-series (TS) signals, while textual information such as news remains largely underutiliz
arXiv:2610.14802v1 Announce Type: new Abstract: End-to-end driving policies can achieve strong task performance while offering limited control over their driving style. Driving style describes how a v
Lauren Forristal / TechCrunch: YouTube Music rolls out labels for songs created with the help of AI; Spotify has AI labels, and Apple Music plans to add “Made With AI” labels later this year — YouTube
arXiv:2610.17521v1 Announce Type: new Abstract: Evaluating VLA policies on real robots is costly, time-consuming, and difficult to reproduce, making real-to-sim simulation essential for scalable evalu
arXiv:2610.15225v1 Announce Type: new Abstract: Zeroth-order (ZO) optimization enables memory-efficient fine-tuning of large language models (LLMs). However, ZO optimization suffers from high variance
arXiv:2610.15336v1 Announce Type: new Abstract: Influence functions enable principled data selection for large language models (e.g., identifying influential training samples for model editing, debugg
arXiv:2610.17696v1 Announce Type: new Abstract: Identifying protein-ligand binding sites (pockets) is important for understanding protein function and developing drugs. However, no public method has l
Here is how i serve 40 people (10 of them developers) with local AI services. n8n for automatisation whisper and Vexa AI, a bot joins Teams and makes transcripts, n8n later extracts decisions and todo
After much testing and talking to actual users across all three proactive AI agents: Muse - best one for your parents, heavy IG/FB users, people who like or need easy setups, people who don’t want to
Follow-up to last week's post about streaming MoE experts from disk in a browser tab. That ran on hand-written engines for two models. This week the same disk tier sits under llama.cpp's own WebGPU ba
Anthropic: Anthropic details Scion, a git replacement for its monorepo, built by seven engineers with Claude in five months, as daily merges rose from 2.5K in May to 30K+ — As our merges per day grew
hexagon: extend scalar DMA broadcasting support for binary ops (#30245) tests: cover weighted scalar broadcasts Assisted-by: OpenCode ggml : support extended scalar DMA binary ops Assisted-by: OpenCod
models : remove redundant ggml_cont calls now supported by all unary kernels (#30198) models: remove redundant ggml_cont calls now supported by CUDA unary kernels models: restore cont_2d after sigmoid
ggml-zdnn : introduce Q8_0 mat mul (#29836) ggml-zdnn: initial prototype of q8_0 matmul Signed-off-by: Aaron Teo aaron.teo1@ibm.com docs: update zdnn docs Signed-off-by: Aaron Teo aaron.teo1@ibm.com S
graph : avoid reallocations in the 'no cache' path (#30439) Website: https://llama.app Attestations: https://github.com/ggml-org/llama.cpp/attestations/55354972 macOS/iOS: macOS Apple Silicon (arm64)
cuda : add GGML_CUDA_GRAPH_DEBUG and read env vars once (#30327) ggml-cuda: gate CUDA graph debug logs behind GGML_CUDA_GRAPH_DEBUG ggml-cuda: cache env var lookups in static const instead of re-readi
vulkan: allow multiple devices to access gpu-visible host memory (#29741) vulkan: allow multiple devices to access gpu-visible host memory clean up erroneous newlines add Windows support handle nullpt
tests : consolidate context state tests into test-llama-context (#30023) tests : consolidate the context state tests into test-llama-context test-save-load-state, test-recurrent-state-rollback and tes
graph: skip the LM head for pooled embeddings (#30445) graph: skip the LM head for pooled embeddings A context that extracts pooled embeddings outputs one vector per sequence, but every model still ex
model: load YaRN beta metadata for laguna models (#30331) Assisted-by: Codex Co-authored-by: Adam Caldwell accaldwell@users.noreply.github.com Website: https://llama.app Attestations: https://github.c
llama : do not map foreign ranges into a backend buffer A buffer created from a host pointer pins everything that it wraps. When several contexts share the same file (lazy tensors, MoE tensors kept on
graph: pad the KQ bias of the no-cache attention (#30446) The no-cache attention pads K, V and the mask to 256, but T5 also adds a relative position bias of [n_tokens, n_tokens], so the add with KQ fa