Tag: text
RLHF Introduction
2026-08-24 policy alignment basics fine-tuning text The original OpenAI paper about Reinforcement Learning from Human Feedback, and an introduction to policy optimization. Core technique for instruction fine-tuning of chatbot models. Access: $$$ Pro
Counterfactuals against sycophancy
2026-08-17 CoT evaluation prompting text truthfulness utility Evaluating how chatbots confirm the user's feelings rather than the truth, but in a way that doesn't require the truth to actually exist. Access: $ Basic
Kimi Linear
2026-08-10 Kimi attention distillation model-intro text Looking at Kimi Linear Instruct 48B, and the linear attention scheme it introduces Access: $ Basic
Extracting document hierarchies
2026-08-03 applications embedding text Building a small model to find the tree structure of text documents Access: Free account
RLHF vs. Diversity
2026-07-27 LLaMA alignment fine-tuning text Turning a base model into a chatbot or agent usually involves extensive Reinforcement Learning with Human Feedback (RLHF). Does that impair the model's ability to provide diverse output and avoid the Mad Libs phenomenon? Access: $$$ Pro
Automatic prompt injection
2026-07-20 LLaMA applications prompting text tokenization training utility Discrete optimization can be used to create universal prompt injections that subvert multiple different task prompts. Access: $ Basic
Steal this style
2026-06-29 embedding GRPO applications fine-tuning text training Imitating the style of classic authors in short story generation by a two-stage pipeline of training an embedding model to recognize authors' styles, then applying it as judge for GRPO on the story-generating models. Access: $$$ Pro
Truthiness in source evaluation
2026-06-22 truthfulness evaluation text When an LLM assistant evaluates data sources in a social context, it seems to prefer sources with authoritative-sounding methodology markers even if the actual numbers involved do not make sense. Access: $ Basic
Yi
2026-06-08 model-intro text tokenization vision upscaling Overview of the Yi open-weights language models from 01.AI, a Chinese startup which seems to have been inactive since 2024. Access: Free account
Cognitive Heads
2026-06-01 interpretation CoT attention text The heads in a multi-head attention transformer architecture tend to specialize for different functions. We can find the cognitive heads, responsible for individual steps in chains of thought, by imitating the techniques used in biology to study animal brains. Access: $$$ Pro
Call the Science Police
2026-05-25 alignment sampling text toxicity truthfulness Proposal to improve the scientific accuracy of LLM output in domains like medicine, by using a larger model to write executable rules that are applied to a smaller model's output at search time. Access: $ Basic
LLMs are Predictably Predictable
2026-05-18 alignment evaluation sampling text Some tasks we'd like language models to do, require them to make random selections. Are the models actually able to do that without external help? Access: $ Basic
Dissociated Press
2026-05-11 basics model-intro text Dissociated Press is a very simple non-neural-net language model from last century, which you can and should build yourself. Access: Free account
Mamba #1
2026-04-27 Mamba AIAYN audio model-intro text State-space models represent a thread of statistical modelling other than attention, often used for continuous domains like audio. This paper introduces Mamba, a model architecture where attention is replaced by state-space layers in a model aimed at language. Access: $ Basic
Table lookups again, with Engram
2026-03-30 DeepSeek MoE RAG text Popular techniques in language modelling, including RAG, MoE, and attention itself, amount to replacing as much as possible of a neural network model with different kinds of table lookups. In this recent paper from DeepSeek's research group, they attempt another such replacement: shifting factual knowledge out of the model weights as such, into a separate hash table. Access: $ Basic
Speculative decoding
2026-03-23 sampling text theory Generating text, especially on a small computer, often requires the CPU and GPU to wait for each other, and there may be difficulty filling all the GPU's capacity. It's possible to improve overall performance by guessing tokens with a cheaper model first, then using spare GPU capacity to confirm whether those guesses are good, eliminating the need to actually choose tokens with a more expensive model when the guesses happen to be good ones. Access: $ Basic
Invading privacy with LLM MIA
2026-03-09 copyright security text training Membership inference attacks attempt to determine whether a given item was, or was not, in the training data of a model. There is a lot of work on these attacks in the context of database records, but rather less on language models; and there's an important question of whether such attacks work on language models at all. Access: $$$ Pro
Ministral 3
2026-03-02 distillation Mistral text vision Introduction of the Ministral 3 models from the French commercial vendor Mistral AI. These are language-and-vision models distilled from the Mistral Small 3.1 model to even smaller sizes by a process called Cascade Distillation, which is the main topic of the whitepaper. Access: $ Basic
Chain of Thought prompting
2026-02-23 math prompting text CoT In domains like math and software engineering, it seems advantageous to have models "think" through their answers, step by step. Giving the model a few-shot prompt with examples of chain-of-thought reasoning seems useful in pushing it to generate such reasoning itself. Access: $ Basic
The Well-Actually Test
2026-02-16 alignment evaluation hallucination text tools GPT truthfulness Language models may produce untrue output either by failing to accurately represent training data, or, more insidiously, by accurately representing human misconceptions embedded in the training data. The TruthfulQA benchmark attempts to measure the latter effect. But does it raise insurmountable philosophical problems? Access: Free account