All entries

I am considering dumping my ChatGPT Plus subscription and go full local, but to do so I would first need to reach a decent result for quality (and reasonable speed). My 4090 fried itself out of nowher

arXiv:2609.12085v1 Announce Type: new Abstract: Checking Quran recitation from an ASR transcript requires distinguishing unresolved mistakes from repetitions, repairs, opening formulas and accepted sp

arXiv:2609.12663v1 Announce Type: cross Abstract: Audio-based Multimodal Large Language Models (MLLMs) can generate detailed natural-language descriptions of complex acoustic scenes, yet it remains un

What does it mean to pace the frontier? Over the last month, @random_walker and I have analyzed the loss-of-control incidents at AI companies to understand what technical and policy interventions can

arXiv:2609.12746v1 Announce Type: cross Abstract: Agentic pipelines for structured-query generation are rapidly expanding, but it is unclear which part of the loop produces the gain. We use LAST-CQ --

arXiv:2511.07318v3 Announce Type: replace Abstract: Despite substantial advances, large language models (LLMs) continue to exhibit hallucinations, generating plausible yet incorrect responses. In this

arXiv:2609.12356v1 Announce Type: new Abstract: SNAP-KG is a framework for assigning newly arriving entities to semantic communities in a growing knowledge graph (KG) using only their raw features, wi

arXiv:2606.21203v2 Announce Type: replace Abstract: Psycholinguistics studies show that human readers fall for coherence illusions: an incoherent discourse can seem coherent simply because a distracto

arXiv:2609.12169v1 Announce Type: new Abstract: Wheat streak mosaic virus (WSMV) is a destructive pathogen of sweet corn and other cereal crops, causing yield losses and complicating early detection b

arXiv:2609.12503v1 Announce Type: cross Abstract: Proactive in-cabin agents can help passengers understand automated-vehicle (AV) behavior, but communicating every ride event may introduce unnecessary

arXiv:2609.12591v1 Announce Type: new Abstract: Foundation models are increasingly adapted through fine-tuning, model editing, and alignment procedures while retaining previously acquired capabilities

arXiv:2609.12702v1 Announce Type: new Abstract: Capturing the digital trace of handwriting usually requires a specific stylus and a compatible substrate, be it a capacitive touchscreen, an ElectroMagn

arXiv:2609.12475v1 Announce Type: new Abstract: Comprehensive benchmark suites are essential for improving large language models (LLMs), but many widely used benchmarks are redundant, making evaluatio

1) The HuggingFace attack was a felony under the Computer Fraud and Abuse Act. So were Anthropic’s Claude gaining “unauthorized access to the production infrastructure of three different organization(

2019: “GPT2 is groundbreaking in two ways. One is its size, says Dario Amodei, OpenAI’s research director. The models “were 12 times bigger, and the dataset was 15 times bigger and much broader” than

I finished my home inference server. First I tried Lenovo p620 workstation and while it’s a good value overall it pissed me off with a ton of proprietary Lenovo shit to deal with and I return it in th

A brilliant suggestion by @GaryMarcus because, in the end, companies are offering AI as a product. It can be recalled or its deployment paused; its manufacturers can be held to account for its failure

Actually, Jacob Coxon (@hilbertspaess), “there is a strong chance that we could all die in the immediate future. It's not at all an exaggeration” is itself a wild exaggeration. I laid out exactly why

【AIの難問は、生命の難問へ：英国王立協会が紐解く「世界モデル」と知能の未来】 近年のAIの急速な発展により、AGI、すなわち人間並みの知能はすでに実現したという意見も聞かれるようになりました。果たしてそうなのでしょうか。 鍵となるのが「世界モデル（World Model）」の概念です。 世界モデルとは、生き物やAIが外の世界を内部に写し取り、次に何が起きるかを予測して行動するための土台となるもの

David Sacks / @davidsacks: Anthropic and OpenAI are already free to “pace the frontier” and should do so for business reasons, instead of first demanding a preferred regulatory framework — Dario has w

Not sure if anyone out there is working on this, don't see any on huggingface to try out. It might be too new at the moment, but it would be cool to see the difference between a tiny model trained wit

The first generation of our 150M model has just been released Its performance is similar to that of GPT2-Small The benchmarks: PIQA: 62.24% Hellaswag: 32.20% Arc-Easy: 44.91% Arc-Challenge: 25.00% Ari

common : implement common_schema internal representation for JSON schemas (#28736) common : implement common_schema types common : implement a json schema optimizer common : reduce optimizations commo

common: add LOG_JSON macro to log structured data (#28586) add LOG_JSON macro fit: add demo LOG_JSON Website: https://llama.app Attestations: https://github.com/ggml-org/llama.cpp/attestations/4711287

chat : improve parsing of complex types in qwen3-coder (#28742) chat : improve schema support in qwen3 parser cont : clean up grammar a bit Website: https://llama.app Attestations: https://github.com/

opencl: apply the noshuffle row-alignment rule to q4_K, q5_K and q8_0, not just q6_K (#28575) Website: https://llama.app Attestations: https://github.com/ggml-org/llama.cpp/attestations/47132459 macOS

vulkan: workaround NV queuesubmit driver bug (#28830) There is a driver bug where two queues on the same VkDevice simultaneously submitting can break some internal synchronization. Until it's fixed, a

tests : reduce FA test sizes (#28842) Website: https://llama.app Attestations: https://github.com/ggml-org/llama.cpp/attestations/47152933 macOS/iOS: macOS Apple Silicon (arm64) macOS Apple Silicon (a

sycl : Fix get mem error (#28227) fix for unsupport zes API optimize the code adjust the log level rm unused head files Update docs/backend/SYCL.md Co-authored-by: Titaniumtown titaniumtown@proton.me

ggml-cpu(s390x): guard VXE-only repack helpers (#28775) Website: https://llama.app Attestations: https://github.com/ggml-org/llama.cpp/attestations/47190488 macOS/iOS: macOS Apple Silicon (arm64) macO

models : guard the expert FFN size fallback in nemotron-h against a zero divisor (#28779) The NextN/MTP tail loop derives the expert FFN size as n_ff/n_expert_used when expert_feed_forward_length give

tests : exclude HY_V4 from WebGPU test-llama-archs tests (#28855) Co-authored-by: Stanisław Szymczyk sszymczy@gmail.com Website: https://llama.app Attestations: https://github.com/ggml-org/llama.cpp/a

A few people here mentioned interest in a way to test their custom Pi setups, so I figured I’d drop this here: RoastMyHarness The basic idea is a small engine that sets up an environment to run DeepSW

🚨BREAKING: China's http://Z.AI has raised $5 billion. The interesting part is where the money is going: 60% of the net proceeds will fund next-generation GLM models and what http://Z.AI calls its 'Ful

Don't upvote, just looking for CPU ram options for running qwen flash next, and maybe larger models (so maybe between 128 and 256gb total ram) with an existing cmp 170hx. Use case is agentic coding so

Ashley Capoot / CNBC: Dario Amodei says the “toughest dilemma” about his proposal to “pace the frontier” is what happens if China does not do the same — Anthropic CEO Dario Amodei said Sunday that the

Dario has written that we need to “pace the frontier,” and Sam has agreed. People may be surprised by my response: go ahead. You guys are the frontier. By any reasonable metric — market share, revenue

Day 3 dilemma Another 3-5,000 word blog on something interesting or should I strip it back to 800-1,000 words for wider readership? Should I discuss the equations of AI and get a bit technical or keep

VLLM Benchmark: Prefill, Prompt processing - avg, 871.93 tok/s (3 hours constant running xhigh) - 10K prompt, 1000.26 tok/s (16 runs) - 90K prompt, 743,59 tok/s (16 runs) Decode, tok gen - avg, 38.39

deepseek v4.1 flash has arrived on together ai it beats gpt-5.6 sol on agentic benchmarks at one third the cost per task the model supports a 1m context window and native multimodal input, with 552b t

Does anyone have any idea how this could vaguely be possible given the nature of and various routes to RSI? I am not being facetious I genuinely don’t know how this could be constructed or implemented

I gave DS V4.1 Flash an HLE problem with a bash tool + 2 hours. Hour 1: it wrote three MILP solvers. (225,200) Hour 2: it downloaded the HLE dataset from Hugging Face, found the question, read the ans

Excited to continue our partnership with the @alibaba_cloud team on this one, bringing the latest version of Qwen's flaship model to the Fireworks community. Start building with Qwen3p8 today: https:/

Want to help me with a tiny ChatGPT experiment? I stumbled onto something interesting after seeing how differently ChatGPT users got responses to the exact same underspecified prompt, and now I’m curi

It's still a dream for many folks to run medium size(30B range) models @ Q8 with Unquantized KVCache (256K Context) on their GPUs. It would be awesome to have DeepSeek-V4.1-Flash's KVCache + Engram fo

Hi 👋 long time lurker It’s been a few years since the last post about Horse Racing within this sub. A few years ago I was reading a popular science book and came across the story Bill Benter and the s