Hacker Papers
- 101
Byte Latent Transformer: Patches Scale Better Than Tokens
Hacker News threads
- May 2025Byte latent transformer: Patches scale better than tokens (2024)
- Dec 2024Tokenisation Is NP-Complete
- Dec 2024Show HN: I made a website to semantically search ArXiv papers
- Feb 2025Ask HN: Is anybody building an alternative transformer?
- May 2025Tokenization for language modeling: BPE vs. Unigram Language Modeling (2020)
- Jun 2026"They're made out of weights"
- Jun 2026Data Compression Explained
- Sep 2026When did Google get so weird?
- Oct 2026Retrofitting language models to operate over bytes
- 102
The Reversal Curse: LLMs trained on "A is B" fail to learn "B is A"
Hacker News threads
- Jun 2026The Reversal Curse: LLMs trained on "A is B" fail to learn "B is A"
- Oct 2023LLMs can't self-correct in reasoning tasks, DeepMind study finds
- Nov 2023Transformers struggle with generalizing tasks beyond pre-training data
- Nov 2023Greg Brockman quits OpenAI
- Jan 2024New theory suggests LLMs can understand text
- Jul 2024Mistral NeMo
- Nov 2025Learn Prolog Now (2006)
- Jul 2026A global workspace in language models
- Aug 2026Tim Gowers: What sort of maths are LLMs good at?
- 103
Transformers are RNNs: Fast Autoregressive Transformers with Linear Attention
Hacker News threads
- Jul 2020Faster Autoregressive Transformers with Linear Attention
- Jul 2020The Reformer – Pushing the limits of language modeling
- Jan 2023What a transformer can NOT do
- Apr 2023There Is No A.I
- Apr 2023There Is No A.I
- Jun 2024Mamba-2 – State Space Duality
- Jul 2024My Python code is a neural network
- Feb 2025LM2: Large Memory Models
- Feb 2026Attention at Constant Cost per Token via Symmetry-Aware Taylor Approximation
- 104
FlashAttention: Fast and Memory-Efficient Exact Attention with IO-Awareness
Hacker News threads
- Jun 2022Flash Attention accelerates GPT2 3x
- Apr 2023Flash Attention
- May 2023FlashAttention: Fast and Memory-Efficient Exact Attention with IO-Awareness
- Dec 2022GPT3/DALL-E2 in Discord, chat like ChatGPT, generate images, and more
- Apr 2023The Coming of Local LLMs
- Apr 2023Stability AI Launches the First of Its StableLM Suite of Language Models
- May 2023
- Dec 2023LLM in a Flash: Efficient LLM Inference with Limited Memory
- Feb 2024LongRoPE: Extending LLM Context Window Beyond 2M Tokens
- Mar 2024Show HN: Flash Attention in ~100 lines of CUDA
- Apr 2025400x AI Acceleration
- 105
Identifying and attacking the saddle point problem in high-dimensional non-convex optimization
Hacker News threads
- Feb 2015Ask HN: How do you find important/interesting new papers
- Dec 2015Machine learning works spectacularly well, but mathematicians aren’t sure why
- Jan 2016New finding: 600M years ago, a single mutation changed everything
- Oct 2017The quest to evolve neural networks through evolutionary algorithms
- Mar 2019Backpropogation Is Just Steepest Descent with Automatic Differentiation (2013)
- Mar 2019Backpropogation Is Just Steepest Descent with Automatic Differentiation (2013)
- Oct 2020Why deep learning works even though it shouldn’t
- Jun 2026Puzzling Success of Overparameterization: Lottery Tickets or Escape Dimensions?
- 106
Improving language models by retrieving from trillions of tokens
Hacker News threads
- Jan 2022WebGPT: Browser-assisted question-answering with human feedback
- Feb 2022AlphaCode as a dog speaking mediocre English
- Jul 2022RETRO is fast
- Dec 2022AlphaCode Attention Visualization
- Dec 2022How Does GPT Obtain Its Ability? Tracing Emergent Abilities of Language Models
- Jan 2023Neural Networks and the Chomsky Hierarchy [DeepMind, 2022]
- May 2023PaLM 2 Technical Report [pdf]
- Nov 2024LLMs have reached a point of diminishing returns
- 107
Large Language Models as Markov Chains
Hacker News threads
- Nov 2024Large Language Models as Markov Chains
- Oct 2024LLMs as Markov Chains
- Oct 2024Large Language Models as Markov Chains
- Jul 2025Problem solving using Markov chains (2007) [pdf]
- Sep 2025The wall confronting large language models
- Oct 2025BERT is just a single text diffusion step
- Nov 2025AI is a front for consolidation of resources and power
- Dec 2025Reflections on AI at the End of 2025
- May 2026Why AI Agents Cannot Change Software Systems
- Jun 2026How LLMs Work
- Sep 2026The Emergent Symbolic Structure of Artificial Neural Networks
- 108
The Internal State of an LLM Knows When It's Lying
Hacker News threads
- Nov 2023The Internal State of an LLM Knows When It's Lying
- Sep 2023The Internal State of a Large Language Model Knows When Its Lying
- May 2023The internal state of an LLM knows when it is lying
- Jun 2023Show HN: StonksGPT – A Natural Language search tool for Stocks and Finance data
- Aug 2023How to Get ChatGPT to Stop Apologizing?
- May 2024Scaling Monosemanticity: Extracting Interpretable Features from Claude 3 Sonnet
- Aug 2024LLMs know more than what they say
- Sep 2024AI is growing faster than companies can secure it, warn industry leaders
- Sep 2024Is telling a model to "not hallucinate" absurd?
- Oct 2024LLMs know more than they show: On the intrinsic representation of hallucinations
- Feb 2026Psychometric Jailbreaks Reveal Internal Conflict in Frontier Models
- 109
Faith and Fate: Limits of Transformers on Compositionality
Hacker News threads
- Apr 2025Faith and Fate: Limits of Transformers on Compositionality
- Jun 2023Faith and Fate: Limits of Transformers on Compositionality
- Jun 2023Faith and Fate: Limits of Transformers on Compositionality
- Jun 2023Weird GPT-4 behavior for the specific string “ davidjl”
- Jun 2023Researchers warn of ‘model collapse’ as AI trains on AI-generated content
- Dec 2023Will scaling work?
- Jan 2024Chess-GPT's Internal World Model
- Jan 2024New theory suggests LLMs can understand text
- Mar 2024Quiet-STaR: Language Models Can Teach Themselves to Think Before Speaking
- Feb 2025Recent results show that LLMs struggle with compositional tasks
- Oct 2025OpenAI researcher announced GPT-5 math breakthrough that never happened
- 110
Driven by Compression Progress: A Simple Principle Explains Essential Aspects of Subjective Beauty, Novelty, Surprise, Interestingness, Attention, Curiosity, Creativity, Art, Science, Music, Jokes
Hacker News threads
- Jan 2017Subjective Beauty, Novelty, Surprise, Interestingness, Attention, Curiosity,
- Aug 2019Driven by Compression Progress (Schmidhuber, 2008)
- Apr 2023Driven By Compression Progress (2008)
- Sep 2017Ask HN: CS as a philosophy
- Oct 2017Intelligence Reconsidered
- Mar 2021Information Is Surprise (2015)
- Mar 2021Hutter Prize
- Jul 2024Creativity fundamentally comes from memorization?
- Mar 2025ARC-AGI without pretraining
- Aug 2026Compression Is Prediction
- Sep 2026Asking Authors About Their Own Papers
- 111
Learning Transferable Visual Models From Natural Language Supervision
Hacker News threads
- Mar 2021The End of Tesla’s Dominance May Be Closer Than It Appears
- Aug 2021Machine learning won't solve natural language understanding
- Apr 2022A few things to try with DALL·E
- Apr 2022How DALL-E 2 Works
- Mar 2024Is Cosine-Similarity of Embeddings Really About Similarity?
- Aug 2024Artists score major win in copyright case against AI art generators
- Jan 2025Don't use cosine similarity carelessly
- Mar 2025Ask HN: Any truly multi-modal transformer architectures?
- 112
High-Resolution Image Synthesis with Latent Diffusion Models
Hacker News threads
- Aug 2022Build Your Own Imagen Text-to-Image Model
- Sep 2022Stable Diffusion forming images from text: image snapshots at each step
- Sep 2022A Web UI for Stable Diffusion
- Mar 2023The Most Cited AI Papers in 2022
- Jul 2023Stanford A.I. Courses
- Feb 2024The VAE Used for Stable Diffusion Is Flawed
- Jul 2024AI models collapse when trained on recursively generated data
- Dec 2025Show HN: I remade my website in the Sith Lord Theme and I hope it's fun
- 113
Monolith: Real Time Recommendation System With Collisionless Embedding Table
Hacker News threads
- Nov 2022Monolith: Real time recommendation system with collisionless embedding table
- Nov 2022TikTok published an article on their recommendation system architecture
- Nov 2022Paper on 'Monolith: Real Time Recommendation System' by Bytedance Inc (TikTok)
- Dec 2022TikTok’s Secret Sauce
- Jan 2023Ask HN: Techwise, how did TikTok get so good?
- Feb 2023The secret sauce of TikTok’s recommendations
- Nov 2023X is changing its algorithm to highlight smaller accounts
- Dec 2024TikTok divestment law upheld by federal appeals court
- Jan 2025TikTok preparing for U.S. shut-off on Sunday
- May 2025You Don't Need Re-Ranking: Understanding the Superlinked Vector Layer
- Feb 2026TikTok's 'Addictive Design' Found to Be Illegal in Europe
- 114
Switch Transformers: Scaling to Trillion Parameter Models with Simple and Efficient Sparsity
Hacker News threads
- Jan 2021Google breaks the trillion-parameter ceiling with the Switch Transformer
- Jan 2021Switch Transformers: Scaling to Trillion Parameter Models with Sparsity
- Jan 2021Switch Transformers: Scaling to Trillion Parameter Models with Simple Sparsity
- Aug 2021Program Synthesis with Large Language Models
- Oct 2021Megatron-Turing NLG 530B, the World’s Largest Generative Language Model
- Jan 2022The Illustrated Retrieval Transformer
- Mar 2023GPT-4
- Jun 2023GPT4 is 8 x 220B params = 1.7T params
- Jul 2023GPT-4 is getting worse over time, not better
- Oct 2023Decomposing Language Models into Understandable Components
- May 2026Granite 4.1: IBM's 8B Model Matching 32B MoE
- 115
Adam: A Method for Stochastic Optimization
Hacker News threads
- Oct 2015Adam's method for stochastic optimization
- Jan 2017A New Batch of Deep Learning Startups to Watch for in 2017
- Oct 2017Variational Inference and Deep Learning: A New Synthesis [pdf]
- Feb 2023Exclusive Q&A John Carmack’s ‘Different Path’ to Artificial General Intelligence
- Jul 2023A Theory on Adam Instability in Large-Scale Machine Learning
- May 2024xLSTM: Extended Long Short-Term Memory
- Oct 2024Anthropic hires OpenAI co-founder Durk Kingma
- Mar 2025The PhD Metagame: Don't try to reform science – not yet
- Oct 2025Researchers Discover the Optimal Way to Optimize
- 116
Binarized Neural Networks: Training Deep Neural Networks with Weights and Activations Constrained to +1 or -1
Hacker News threads
- Sep 2016Stacked Approximated Regression Machine: A Simple Deep Learning Approach
- Oct 2016Intel will add deep-learning instructions to its processors
- Feb 2017Intel Gets Serious About Neuromorphic, Cognitive Computing Future
- Sep 2017Nonlinear Computation in Deep Linear Networks
- Oct 2018MIT AGI: Conversation with Yoshua Bengio [video]
- Nov 2019Deep Learning and Its Implications for Computer Architecture and Chip Design
- Feb 2024The Era of 1-bit LLMs: ternary parameters for cost-effective computing
- Mar 2024Towards 1-bit Machine Learning Models
- 117
Neural Networks and the Chomsky Hierarchy
Hacker News threads
- Jul 2022Neural Networks and the Chomsky Hierarchy by DeepMind
- Jan 2023Neural Networks and the Chomsky Hierarchy [DeepMind, 2022]
- Aug 2022Deep Learning Alone Isn’t Getting Us to Human-Like AI
- Feb 2023What Is ChatGPT Doing and Why Does It Work?
- Apr 2023Tighter bounds on the expressivity of transformer encoders
- May 2023Ask HN: Why can GPT4 en/decode base64?
- Jun 2023Bard is getting better at logic and reasoning
- Jun 2023Modern language models refute Chomsky’s approach to language
- Nov 2024Interview with gwern
- Jun 2026No, Artificial Intelligence is not Conscious - Ted Chiang
- 118
Memory Augmented Large Language Models are Computationally Universal
Hacker News threads
- Jul 2023Memory Augmented Large Language Models Are Computationally Universal
- Feb 2023Memory Augmented Large Language Models Are Computationally Universal
- Mar 2023Memory Augmented Large Language Models Are Computationally Universal
- Feb 2023Our Society Is Not Prepared for This Much AI Awesome
- Mar 2023Show HN: Whatdoesthiscodedo.com – AI explanations for other people’s code
- Apr 2023Teaching ChatGPT to Speak My Son’s Invented Language
- Apr 2023Large Language Models: Scaling Laws and Emergent Properties
- May 2023Anthropic - Introducing 100K Token Context Windows, Around 75,000 Words
- Oct 2023Ask HN: Why are current LLMs not considered AGI?
- Sep 2025Markov chains are the original language models
- May 2026Claude Opus 4.8
- 119
Twitter mood predicts the stock market
Hacker News threads
- May 2011Twitter mood predicts the stock market.
- Nov 2010Twitter mood predicts the stock market
- May 2011Twitter mood predicts the stock market
- Oct 2010Twitter Mood Predicts The Stock Market
- Nov 2010Can LinkedIn predict when companies are in trouble?
- Dec 2010Hedge fund to use Twitter to predict stock market
- Jul 2011MongoDB is the New MySQL
- Mar 2012What you could do if you were Google and had their databases
- May 2012Looking At The World Through Twitter Data
- Sep 2012Twitter is a tragic tale
- Mar 2014Can the Bloomberg Terminal be "Toppled"?
- 120
Practical Bayesian Optimization of Machine Learning Algorithms
Hacker News threads
- Apr 2015Practical Bayesian Optimization of Machine Learning Algorithms (2012)
- May 2015PRACTICAL BAYESIAN OPTIMIZATION OF MACHINE LEARNING ALGORITHMS [pdf]
- Jul 2012Machine learning for the impatient: algorithms tuning algorithms
- Nov 2012How deep learning on GPUs wins datamining contest without feature engineering
- Jan 2013Ersatz - Deep neural networks in the cloud
- Nov 2013Machine learning is easier than it looks
- Dec 2013Hyperopt: Optimize your machine learning parameters automagically
- Jun 2015Whetlab.com: API for Baysian optimization of hyperparameters
- Jan 2018Show HN: A radically more scalable alternative to Bayesian Optimization
- Jul 2018Troubling Trends in Machine Learning Scholarship
- 121
Man is to Computer Programmer as Woman is to Homemaker? Debiasing Word Embeddings
Hacker News threads
- Jul 2016Debiasing Word Embeddings
- Jul 2016Man Is to Computer Programmer as Woman Is to Homemaker? Debiasing Word Embedding
- Jan 2017Man to Computer Programmer as Woman is to Homemaker? Debiasing Word Embeddings
- Sep 2016A.I. Doesn't Get Black Twitter
- Jul 2017Ask HN: Any tools to easily cluster ideas/notes by similarity?
- Oct 2017Copyright Law Can Fix Artificial Intelligence’s Implicit Bias Problem
- Mar 2018Show HN: Word2Bits – Quantized Word Vectors
- Oct 2018Show HN: Basilica – word2vec for anything
- Jul 2019King – Man + Woman = King?
- Sep 2019600k Images Removed from ImageNet After Art Project Exposes Racist Bias
- Sep 2020FB to curb internal debate over sensitive issues amid employee discord
- 122
The Loss Surfaces of Multilayer Networks
Hacker News threads
- Jan 2019The Loss Surfaces of Multilayer Networks
- Mar 2015Deep Learning vs. Machine Learning vs. Pattern Recognition
- Jun 2015What's Wrong with Deep Learning?
- Aug 2015Machine Learning: Curse of Dimensionality
- Feb 2016AI is transforming Google search – the rest of the web is next
- Aug 2016What’s Next for Artificial Intelligence
- May 2017Breaking Glass in Infinite Dimensions
- Dec 2017An Interview with an Anonymous Data Scientist (2016)
- Aug 2020Differentiable Control Problems
- 123
ByT5: Towards a token-free future with pre-trained byte-to-byte models
Hacker News threads
- May 2022Teaching GPT-3 to reverse words
- Dec 2022ChatGPT, Rot13, and Daniel Kahneman
- Mar 2023Cerebras-GPT vs. LLaMA AI Model Performance Comparison
- Apr 2023A Baby GPT
- Jun 2023Understanding GPT Tokenizers
- Aug 2024Markov chains are funnier than LLMs
- Nov 2024Something weird is happening with LLMs and chess
- Jun 2026Data Compression Explained
- 124
GSM-Symbolic: Understanding the Limitations of Mathematical Reasoning in Large Language Models
Hacker News threads
- Oct 2024Understanding the Limitations of Mathematical Reasoning in LLMs
- Oct 2024Cwal: Scripting Engine Without a Language
- Oct 2024Language is not essential for the cognitive processes that underlie thought
- Oct 2024OpenAI builds first chip with Broadcom and TSMC, scales back foundry ambition
- Nov 2024OK, I can partly explain the LLM chess weirdness now
- Dec 2024GPT-5 is behind schedule
- Jan 202530% drop in O1-preview accuracy when Putnam problems are slightly variated
- Jun 2025The Illusion of Thinking: Strengths and limitations of reasoning models [pdf]
- Jul 2026Emily Bender Sets the Record Straight on "Stochastic Parrots"
- 125
Extracting books from production language models
Hacker News threads
- Jan 2026Extracting books from production language models
- Jan 2026Extracting books from production language models
- Jan 2026Extracting books from production language models (2026)
- Jan 2026Nvidia Contacted Anna's Archive to Access Books
- Feb 2026We tasked Opus 4.6 using agent teams to build a C Compiler
- Feb 2026Microsoft offers guide to pirating Harry Potter series for LLM training
- Mar 2026Relicensing with AI-Assisted Rewrite
- Mar 2026FSF Threatens Anthropic over Infringed Copyright: Share Your LLMs Freel
- Jun 2026Did Claude Increase Bugs in Rsync?
- Jul 2026Judge approves $1.5B Anthropic settlement for pirated books used to train Claude
- Sep 2026When did Google get so weird?
Data from the Hacker News archive (ODC-By), arXiv and OpenAlex. Not affiliated with Hacker News or Y Combinator.