imudojuiwọn ojoojumọ3981 daju itan
AI iroyin. Laisi ariwo.
Agbegbe AI ti a ṣayẹwo orisun ti awọn ifilọlẹ ọja, awọn iyipada eto imulo, iwadii aabo, ati awọn gbigbe ile-iṣẹ, ṣalaye ni ede Gẹẹsi nipasẹ ẹgbẹ ẹkọ ti kii ṣe èrè.
Ifowosowopo ti o ni idaniloju
Gbogbo itan ni asopọ si ẹri ti o lagbara julọ ti o wa: awọn orisun atilẹba nigbati o wa, bibẹkọ ti o han gbangba.
English itele
Kini o ṣẹlẹ, idi ti o ṣe pataki, ati kini lati wo - laisi jargon.
Ko si kikun
Nigbati awọn ifihan agbara jẹ tinrin, a jade ohunkohun kuku ju òwú kikọ sii.
Awọn itan diẹ sii
9 awọn itanAtunse
MoPLEx proposes a way to model mixed human preferences in AI alignment
A new arXiv preprint proposes MoPLEx, an algorithm for learning from multi-way rankings when annotators have different underlying preferences. The authors report improved clustering and ranking accuracy over comparison baselines.arxiv.orgAtunse
Awọn igbewọle ti ara-alaiyipada ti ara fẹẹrẹ di ìlọpo mẹta awoṣe ede-iṣawari PDE-ipejuwe, iṣaju iṣaju sọ
Atejade arXiv kan nipasẹ Fan Yang ati Matt Thomson ni imọran titumọ data aaye aaye aye sinu awọn iwọn ti ara ṣaaju fifun ni awoṣe ede nla kan. Lori awọn ipilẹ aaye ti afọwọṣe, awọn onkọwe sọ pe ipele itumọ yii ti fẹrẹ to ilọpo mẹta ti a gba pada-idogba ni ibamu pẹlu fifi data aise awoṣe han…arxiv.orgAtunse
Preprint argues LLM quantization needs transforms designed for the number format
An arXiv preprint surveys transform methods used before quantizing large language models and identifies a conflict between classical transform coding and grouped, shared-scale quantization. It argues that the best transform depends jointly on the deployment format and how scale information is allocated.arxiv.orgAtunse
Study finds attention architecture drives the energy cost of LLM inference
A new empirical study reports that attention architecture strongly affects how large language model inference energy grows with context length. It finds steeper growth for multi-head attention, flatter growth for grouped-query attention, and nearly constant energy for grouped-query attention combined with…arxiv.orgAtunse
Researchers add a quantization layer to AI-based radio interference suppression
A new arXiv paper describes an AI system that combines an autoregressive transformer with a Finite Scalar Quantization tokenizer to suppress structured radio-frequency interference. The authors report lower latency and stronger interference rejection than traditional and earlier AI-based approaches, but provide no…arxiv.orgAtunse
Paper proposes lower-cost pruning for Transformer models during fine-tuning
REP-LIE estimates which Transformer weights to remove using gradients from LoRA low-rank matrices, aiming to reduce the resources required for model pruning and subsequent fine-tuning.arxiv.orgAtunse
Iwadi n ṣe idanimọ nigbati awọn alaiṣẹ-nẹtiwọọki neural ṣe ilọsiwaju iṣapeye
Akọsilẹ arXiv tuntun kan jiyan pe awọn onibajẹ-nẹtiwọọki-nẹtiwọọki ṣe iranlọwọ iṣapeye nikan nigbati wọn ṣe iranlọwọ yiyan oludije laarin agbegbe ti o ni opin ati fi ẹrọ gbigba ti o wa labẹ imule.arxiv.orgAtunse
AFDBench tests whether AI can write National Weather Service forecast discussions from weather data
A new preprint introduces AFDBench, a benchmark for testing whether language models can turn structured AI weather forecasts into accurate, professionally written National Weather Service discussions.arxiv.orgAtunse
Preprint explains what reinforcement learning changes inside language models
A new preprint breaks down reinforcement-learning post-training for language models, examining how rewards, prompts, model scale and prior behavior shape the results.arxiv.org
Alaye ti o wulo ni ọsẹ kọọkan
Tẹsiwaju pẹlu AI laisi gbigbe ninu ifunni.
Gba awọn iroyin AI ti a ṣayẹwo ti ọsẹ naa, data atilẹba, awọn irinṣẹ to wulo, awọn yiyan ẹkọ, ati awọn iṣẹ AI tuntun.
De ọdọ awọn eniyan ti o nkọ AI
Igbanisise ọjọgbọn AI kan tabi ṣe ifilọlẹ ọja AI ti o wulo? Fi si iwaju awọn eniyan ti o wa nibi lati kọ ẹkọ ati ṣe.
Firanṣẹ iṣẹ AI kanFi ohun elo AI silẹ