Hacker NewsI built a tiny LLM to demystify how language models work
https://github.com/arman-bd/guppylm
#HackerNews #tinyLLM #languageModels #AI #demystify #coding #guppylm
Emotion concepts and their function in a large language model
https://www.anthropic.com/research/emotion-concepts-function
#HackerNews #EmotionConcepts #LanguageModels #AIResearch #NaturalLanguageProcessing #MachineLearning
sayzardfly51fly (@fly51fly)
대형 언어모델이 가르치는 과정에서 상대방의 정신 상태를 추론하는지(mentalize) 분석한 2026년 논문이 공개되었습니다. 인간의 teaching 상호작용을 모사하는 LLM의 인지적 행동을 다뤄, 모델 해석과 인간유사 추론 능력 연구에 의미가 있습니다.
Marcus SchulerImplicator.ai released the AI Top 40, a weekly ranking that combines 10 benchmarks into one score per language model. The system weights contamination-resistant tests like SWE-bench 4x higher than Chatbot Arena. GPT-5.4 currently leads despite Claude topping Arena rankings. Updates every Saturday and offers free embedding for websites.
🌎 Experiencia interdimensionalA new study explores how human confidence in large language models (LLMs) often surpasses their actual accuracy. It highlights the 'calibration gap' - the difference between what LLMs know and what users think they know.
Read Full Article
#LanguageModels #AIConfidence #CalibrationGap #MachineLearning #DataAccuracy https://doi.org/10.1038/s42256-024-00976-7
Reenviado desde Science News
(https://t.me/experienciainterdimensional/10502)
Read Full Article
#LanguageModels #AIConfidence #CalibrationGap #MachineLearning #DataAccuracy https://doi.org/10.1038/s42256-024-00976-7
Reenviado desde Science News
(https://t.me/experienciainterdimensional/10502)
Simplifying AI (@simplifyinAI)
Tencent과 Tsinghua가 CALM(Continuous Autoregressive Language Models)을 공개했습니다. 기존 LLM의 next-token 예측 패러다임을 대체하는 방식으로, 이산적 단일 토큰 예측에 쓰이던 막대한 연산 낭비를 줄이는 새로운 언어모델 접근법을 제시합니다.
Simplifying AI (@simplifyinAI) on X
🚨 BREAKING: Tencent has killed the “next-token” paradigm. Tencent and Tsinghua has released CALM (Continuous Autoregressive Language Models), and it completely disrupts the next-token paradigm. LLMs currently waste massive amounts of compute predicting discrete, single tokens
Leon FurzeTeaching Writing in the Age of AI: Assessment and “Cheating”
This is the fourth post in a series on Teaching Writing in the Age of AI. The first post provided an overview of some of the changes we're facing as the number of AI writing tools increases. Post two covered conversations about academic integrity, and the third post offered some practical advice on teaching students to be critical readers and writers. In this post, I'll be exploring the assessment of writing, and why AI is such an apparent threat to the way we currently teach and assess. In […]https://leonfurze.com/2023/02/18/teaching-writing-in-the-age-of-ai-assessment-and-cheating/
N-gated Hacker NewsIn a groundbreaking revelation that nobody asked for, two authors claim language models are actually just like distributed systems... because, you know, both involve computers and math 🤯. They proceed to drown us in a sea of jargon and acronyms, leaving readers wondering if this paper is an elaborate AI-generated prank 😜.
https://arxiv.org/abs/2603.12229 #languagemodels #distributedsystems #AIresearch #jargonoverload #techhumor #HackerNews #ngated
https://arxiv.org/abs/2603.12229 #languagemodels #distributedsystems #AIresearch #jargonoverload #techhumor #HackerNews #ngated
Language Model Teams as Distributed Systems
Large language models (LLMs) are growing increasingly capable, prompting recent interest in LLM teams. Yet, despite increased deployment of LLM teams at scale, we lack a principled framework for addressing key questions such as when a team is helpful, how many agents to use, how structure impacts performance -- and whether a team is better than a single agent. Rather than designing and testing these possibilities through trial-and-error, we propose using distributed systems as a principled foundation for creating and evaluating LLM teams. We find that many of the fundamental advantages and challenges studied in distributed computing also arise in LLM teams, highlighting the rich practical insights that can come from the cross-talk of these two fields of study.
Ah, the thrilling art of "writing" software by letting Large Language Models do all the heavy lifting 🤖💪. Because who needs the joy of coding when you can sit back, sip coffee, and let AI run wild while you take all the credit? 🚀☕
https://www.stavros.io/posts/how-i-write-software-with-llms/ #AIsoftware #CodingJoy #Automation #ThrillingTech #LanguageModels #HackerNews #ngated
https://www.stavros.io/posts/how-i-write-software-with-llms/ #AIsoftware #CodingJoy #Automation #ThrillingTech #LanguageModels #HackerNews #ngated
🤯 Ah, the life of a tortured soul smothered by the oppressive weight of Large Language Models. Who knew that conversing with AI could be so *exhausting* that it saps the energy of our brave hero after a mere 45-hour "sprint"? 😴 Sounds like someone needs a nap or at least a better excuse for those "degrading" prompts. 🙄
https://tomjohnell.com/llms-can-be-absolutely-exhausting/ #torturedsoul #AIexhaustion #languagemodels #naptime #degradingprompts #HackerNews #ngated
https://tomjohnell.com/llms-can-be-absolutely-exhausting/ #torturedsoul #AIexhaustion #languagemodels #naptime #degradingprompts #HackerNews #ngated
