b11034
vocab : add ufakzeka pre-tokenizer (#29033) vocab : add ufakzeka pre-tokenizer vocab : move ufakzeka to the models list and regenerate the hash mapping Website: https://llama.app Attestations: https:/
Knowledge catalogue
vocab : add ufakzeka pre-tokenizer (#29033) vocab : add ufakzeka pre-tokenizer vocab : move ufakzeka to the models list and regenerate the hash mapping Website: https://llama.app Attestations: https:/
vulkan: add IQ3_S MMQ matmul kernels (#28822) vulkan: add IQ3_S MMQ matmul kernels Make block_a_to_shmem do 2-byte loads (110 bytes is divisible by 2) Align the check, IQ3_S is also using K tile size
ggml : handle graph buffer reservation failure (#26070) Website: https://llama.app Attestations: https://github.com/ggml-org/llama.cpp/attestations/48431150 macOS/iOS: macOS Apple Silicon (arm64) macO
ggml-webgpu: fix supports_op condition for GET_ROWS (#28978) fix get_rows vec4 handling Add src strides checking to vec4_aligned of get_rows and the new test case. Website: https://llama.app Attestati
Model-Saver: Write the SWA pattern, 15 more architectures roundtrip (#29042) llama: read the SWA pattern as a period or a per-layer array Add llama_model_base::load_swa_pattern(), which reads sliding_
ggml : check for allocation failures to prevent crashes (#28149) ggml : check for allocation failures to prevent crashes wording Website: https://llama.app Attestations: https://github.com/ggml-org/ll
opencl: add bin kernel kernel_gemm_noshuffle_q6_k_f32_32b_trans_ila_a8_bin (#28678) opencl: add A8 Q6_K non-MoE binary kernel opencl: fix layout compatibility Website: https://llama.app Attestations:
hexagon: HMX flash-attention head_dim padding (support DK=DV=72) (#26539) Allow HMX flash-attention to run with head_dim not a multiple of 64 (e.g. SigLIP head_dim=72), by operating on DK/DV rounded u
hexagon: im2col update (#29103) ggml-hexagon: accept 1D and padded IM2COL ops ggml-hexagon: make pure-DDR IM2COL kernel is_2D-aware ggml-hexagon: extend IM2COL DMA patch-embed fast path to 1D ggml-hex
hexagon: add ROLL op support (#29105) Website: https://llama.app Attestations: https://github.com/ggml-org/llama.cpp/attestations/48565469 macOS/iOS: macOS Apple Silicon (arm64) macOS Apple Silicon (a
opencl: add support for bin kernel flash_attn_f32_f16_bin (#29046) opencl: add flash_attn_f32_f16_bin opencl: guarded prefill fa Website: https://llama.app Attestations: https://github.com/ggml-org/ll
arXiv:2609.20223v1 Announce Type: new Abstract: We study weakly supervised incremental telecom fraud detection from raw telephone audio, where training provides only conversation-level labels and pred
arXiv:2609.19989v1 Announce Type: new Abstract: The widespread adoption of LLMs has led to escalating content compliance risks. Prior works have contributed to addressing these risks in the English co
arXiv:2609.19767v1 Announce Type: new Abstract: True machine intelligence requires transcending passive pixel registration to master top-down functional reasoning over absent information via visual ne
arXiv:2609.19934v1 Announce Type: new Abstract: Depth-recurrent language models iteratively apply a small layer stack, decoupling per-token compute from distinct parameter count. To determine whether
arXiv:2609.19842v1 Announce Type: new Abstract: Effective EEG decoding requires representations that preserve organization among channels, local waveform dynamics, and long-range temporal context. Exi
arXiv:2609.19966v1 Announce Type: new Abstract: Synthetic image and mask pairs can alleviate scarce colonoscopy annotations, but realistic synthesis requires preserving the supplied lesion while gener
Big fan of how @LangChain is doing these insta-webinars The news is a few days old, @typesafeai is integrated and documented with example code in LangChains Docs 👇 And they are doing a webinar on the
arXiv:2609.19180v1 Announce Type: new Abstract: Language models face unique challenges in analyzing interdisciplinary scientific research literature. In biophysics research, faithful answers require g
arXiv:2602.16870v2 Announce Type: replace Abstract: The Boreas Road Trip (Boreas-RT) dataset extends the multi-season Boreas dataset to new and diverse locations that pose challenges for modern autono
Build your own harness, folks. This is absolute banger paper from NVIDIA on self-evolving agent harnesses. (bookmark it) They introduce SoL-Pi which cuts token traffic by nearly half. And it matches i
arXiv:2509.25248v2 Announce Type: replace-cross Abstract: Automatically compiling open-source software (OSS) projects is a vital, labor-intensive, and complex task, which makes it a good challenge for
arXiv:2609.19354v1 Announce Type: cross Abstract: Automated action quality assessment (AQA) in Olympic sports remains a challenging task due to the complexity of human motion and the subjectivity inhe
arXiv:2609.19606v1 Announce Type: new Abstract: This empirical study is an independent reproduction of the dissociation Zhao reported in 2026. The shape of a large language model's chain-of-thought en
AI is accelerating software development at an unprecedented pace. But as code generation scales, so do the challenges of securing the code, especially emerging AI-based vulnerability exploitations. To
arXiv:2609.19244v1 Announce Type: new Abstract: Conversational LLM agents increasingly rely on Web search, yet the end-to-end lifecycle of agentic search remains poorly understood. We present the firs
arXiv:2609.20625v1 Announce Type: cross Abstract: Large language model responses are non-deterministic, so failures in LLM agents are hard to reproduce: a failure depends on inference that is not bitw
arXiv:2506.17277v2 Announce Type: replace-cross Abstract: The retrieval stage of retrieval-augmented generation (RAG) for scientific question answering depends on how documents are segmented and how c
arXiv:2609.19911v1 Announce Type: new Abstract: 3D grounding aims to localize target entities in complex scenes from natural language and plays a fundamental role in embodied perception and spatial re
arXiv:2609.19892v1 Announce Type: cross Abstract: As agent systems become more widely used, multiple agent sessions increasingly run alongside pre-existing user tasks in the same environment, sharing
arXiv:2609.19585v1 Announce Type: cross Abstract: In mental health care, reasoning over patient journeys is a key task for clinicians. Yet these journeys, encompassing a longitudinal progression of bi
arXiv:2609.19425v1 Announce Type: new Abstract: Tool-augmented large language model (LLM) agents fail in a way no tool-selection or tool-security method addresses: they call tools that do not exist an
arXiv:2609.19923v1 Announce Type: new Abstract: Vision-language-action models (VLAs) have emerged as a promising paradigm for general-purpose robot learning, with performance improving as models and d
arXiv:2609.20386v1 Announce Type: new Abstract: Iris presentation attack detection (PAD) is security-critical when a subsystem that appears reliable during development encounters presentation attack i
arXiv:2606.13028v2 Announce Type: replace-cross Abstract: Depth estimation has numerous medical and surgical applications. We benchmark four depth sensors on a porcine bone specimen, a porcine belly s
arXiv:2509.09121v2 Announce Type: replace Abstract: Large language models (LLMs) excel in general-domain applications, yet their performance often degrades in specialized tasks requiring domain-specif
arXiv:2609.19539v1 Announce Type: cross Abstract: Active subspace methods provide a framework for quantifying predictive uncertainty in high-dimensional models by identifying and performing inference
arXiv:2609.19525v1 Announce Type: cross Abstract: Clinical B-mode images are widely available as potential data sources for quantitative ultrasound (QUS) analysis for tissue characterization. However,
arXiv:2609.19674v1 Announce Type: new Abstract: A learned simulator can reproduce its training conditions accurately yet fail in two distinct ways once those conditions change. Over long rollouts, sma
arXiv:2609.19789v1 Announce Type: new Abstract: Multi-agent trading systems built on large language models (LLMs) are beginning to appear in quantitative finance, yet their robustness to adversarial i
arXiv:2609.20586v1 Announce Type: cross Abstract: Referring scene understanding for embodied robots requires grounding object- and relation-centric language queries from a designated viewpoint. While
arXiv:2609.19818v1 Announce Type: cross Abstract: Generalizing to unseen attacks remains challenging for audio deepfake detectors, and collecting training data covering all potential attacks is imprac
arXiv:2609.20045v1 Announce Type: cross Abstract: A memory can answer a current query correctly while discarding distinctions required by a later update. We investigate this failure with a paired-hist
arXiv:2609.20461v1 Announce Type: cross Abstract: Studying the dynamical behavior of a system often depends on characterizing how it transitions between long-lived states. Because such transitions are
arXiv:2609.20260v1 Announce Type: cross Abstract: We comment on the optimally-transported generalized method of moments (OTGMM) estimator proposed by Schennach & Starck (2026a) and give counterexample
arXiv:2609.20441v1 Announce Type: new Abstract: Geospatial foundation models can provide strong flood-segmentation performance, but their size limits deployment on memory-constrained edge hardware. We
arXiv:2609.20592v1 Announce Type: new Abstract: Crystal structure refinement is a fundamental inverse problem in materials characterization, where structural parameters are optimized to reproduce expe
So I know that Ollama is built on top of llama.cpp, I decided to run a model on both and applied settings on llama.cpp like flash attention and q8/q4 kv cache. On Ollama (using the Python package) I w
Three cybersecurity researchers used Claude to breach OpenAI Group PBC’s GitHub repository. Sources told the Wall Street Journal today that the repository contains “OpenAI’s algorithmic secrets.” The
arXiv:2609.20051v1 Announce Type: new Abstract: Step distillation reduces the cost of video generation, but reusing a LoRA trained for a longer trajectory can alter its functional effect or degrade ta
Hi there, I'm Yuesong, a researcher at Video Rebirth. We open-sourced HyperFlow, an 8-step LoRA for MiniMax-H3. The core idea: data-free flow self-distillation. Diffusers' default sampler does 49 mode
arXiv:2609.19969v1 Announce Type: new Abstract: The widespread adoption of long-horizon agents has made model workloads increasingly input-heavy. Although prior work has substantially reduced the cost
I just released a new dataset https://huggingface.co/datasets/LocalLLaMA/deepswe-mini it is meant to be more efficent when testing and benchmarking local models on DeepSwe https://deepswe.datacurve.ai
arXiv:2609.19607v1 Announce Type: cross Abstract: Coding-agent benchmarks are built for broad and comprehensive comparisons, not frequent development decisions. Individual runs vary, full suites are e
arXiv:2609.20104v1 Announce Type: new Abstract: We describe the architecture, training methodology and inference speedups of Granite 5.0 Turbo CTC, a 470 million parameter encoder-only model with an e
arXiv:2609.19805v1 Announce Type: new Abstract: Grapheme-to-phoneme (G2P) conversion turns raw text into its phonemic form and is an essential part of both text-to-speech (TTS) and automatic speech re
arXiv:2609.19913v1 Announce Type: new Abstract: The study of opinion dynamics in social networks is one of the key challenges in computational social science with direct relevance to understanding pol
arXiv:2609.20574v1 Announce Type: new Abstract: Answer grounding in document visual question answering remains an open challenge: most benchmarks lack grounding annotations or provide limited-quality
arXiv:2603.07998v2 Announce Type: replace-cross Abstract: Aerodynamic promptness quantifies how rotor-speed variations generate multirotor wrench variations, but its Euclidean formulation assigns the
Note not my work, but something i found and wanted to share so hopefully more people can push this along even further. https://github.com/nasone32/llama.cpp-RDNA3-7900xtx-opt I basically run 2 x 7900