#language_models — Public Fediverse posts

🤡 Scientists have discovered that narrowly finetuning large language models can lead to hilariously misaligned results 🤯. Who knew that stretching a rubber band in one place would make the whole thing snap? 🙄 Bravo to the geniuses who spend years fine-tuning #chaos. 👏
https://arxiv.org/abs/2502.17424 #scientificdiscovery #humor #language_models #misalignment #fine_tuning #HackerNews #ngated

"🧐 Researchers bravely attempt to 'liberate' snippets from books using language models, ignoring copyright like it's an optional suggestion. 📚🤖 Meanwhile, #arXiv is casually looking to hire a #DevOps engineer, because who doesn't want to work for a glorified PDF repository? 💻🎉"
https://arxiv.org/abs/2505.12546 #liberationofknowledge #copyrightissues #language_models #hiring #HackerNews #ngated

Block Diffusion: Interpolating Autoregressive and Diffusion Language Models
https://m-arriola.com/bd3lms/
#ycombinator #block_diffusion #discrete #masked #diffusion #language_models #BD3_LM #BD3_LMs

Ah, the riveting world of "circuit tracing" in language models 🤖🔍, because what we really needed was another way to complicate things we barely understand. A "replacement model" that makes things "interpretable"? 😂 More like a desperate attempt to justify endless AI research grants.
https://transformer-circuits.pub/2025/attribution-graphs/methods.html #circuittracing #AIinterpretability #researchgrants #language_models #techhumor #HackerNews #ngated

Ah, the riveting world of "circuit tracing" in language models 🤖🔍, because what we really needed was another way to complicate things we barely understand. A "replacement model" that makes things "interpretable"? 😂 More like a desperate attempt to justify endless AI research grants.
https://transformer-circuits.pub/2025/attribution-graphs/methods.html #circuittracing #AIinterpretability #researchgrants #language_models #techhumor #HackerNews #ngated

Ah, the riveting world of "circuit tracing" in language models 🤖🔍, because what we really needed was another way to complicate things we barely understand. A "replacement model" that makes things "interpretable"? 😂 More like a desperate attempt to justify endless AI research grants.
https://transformer-circuits.pub/2025/attribution-graphs/methods.html #circuittracing #AIinterpretability #researchgrants #language_models #techhumor #HackerNews #ngated

Ah, the riveting world of "circuit tracing" in language models 🤖🔍, because what we really needed was another way to complicate things we barely understand. A "replacement model" that makes things "interpretable"? 😂 More like a desperate attempt to justify endless AI research grants.
https://transformer-circuits.pub/2025/attribution-graphs/methods.html #circuittracing #AIinterpretability #researchgrants #language_models #techhumor #HackerNews #ngated

ChatGPT Learned to Reason [video]
https://www.youtube.com/watch?v=PvDaPeQjxOE
#ycombinator #AI_reasoning #ChatGPT_explained #artificial_intelligence #neural_networks #Monte_Carlo_Tree_Search #DeepMind #AlphaGo #chess_AI #language_models #machine_learning #reinforcement_learning #deep_learning #AI_history #GPT_training #chain_of_thought #AI_breakthrough #game_AI #TD_Gammon #MuZero #Claude_AI #O1_AI #AI_algorithms #AI_development #computer_reasoning #AI_evolution #future_AI

Kolmogorov-Arnold Networks: MLP vs. Kan, Math, Universal Approximation Theorem [video]
https://www.youtube.com/watch?v=-PFIkkwWdnM
#ycombinator #pytorch #python #tutorial #math #language_models #deep_learning #machine_learning #multi_layer_perceptron #mlp #kolmogorov_arnold_networks #kolmogorov_arnold_representation_theorem #universal_approximation_theorem #neural_networks #bezier_curves #splines #b_splines #linear_layers