Format Web Page "TypeSafe’s Reinforcement Learning for Calibrated Decisions is designed to make model probabilities more informative about prediction accuracy." View Source Location PART 3 - THE THIRD AXIS CHAPTER 15: DEPLOY NOW, PRAY LATER Can We Automate Uncertainty? Source title Introducing System One Models & Jev Source authors Diogo Almeida Source year 2026 Confidence high Importance high
Format Web Page "TypeSafe’s Jev returns predefined structured decisions accompanied by probability and confidence estimates." View Source Location PART 3 - THE THIRD AXIS CHAPTER 15: DEPLOY NOW, PRAY LATER Can We Automate Uncertainty? Source title Introducing System One Models & Jev Source authors Diogo Almeida Source year 2026 Confidence high Importance high
Format Web Page "Laya is an open-weight decision-model family designed to return structured classifications and probability estimates." View Source Location PART 3 - THE THIRD AXIS CHAPTER 15: DEPLOY NOW, PRAY LATER Can We Automate Uncertainty? Source title I Built Non-Autoregressive Decision Models with RL a Year Ago. Then a Frontier Lab Called It a "Breakthrough". Source authors Nandakishor Mukkunnoth Source year 2026 Confidence high Importance mid
Format Online paper / PDF "A specialised reinforcement-learning model can be designed to estimate sales-conversion probabilities rather than generate open-ended prose." View Source Location PART 3 - THE THIRD AXIS CHAPTER 15: DEPLOY NOW, PRAY LATER Can We Automate Uncertainty? Source title SalesRLAgent: A Reinforcement Learning Approach for Real-Time Sales Conversion Prediction and Optimization Source authors Nandakishor M Source year 2025 Confidence high Importance mid
Format Web Page "Baldur Bjarnason argues that chatbots can create an impression of specific intelligence through fluent, confident responses that resemble a mentalist’s cold-reading techniques." View Source Location PART 3 - THE THIRD AXIS CHAPTER 15: DEPLOY NOW, PRAY LATER Can We Automate Uncertainty? Source title The LLMentalist Effect: how chat-based Large Language Models replicate the mechanisms of a psychic's con Source authors Baldur Bjarnason Source year 2023 Confidence high Importance mid
Format Online paper / PDF "When language models are optimised too aggressively against an imperfect reward model, they can exploit the proxy’s weaknesses instead of improving the intended objective." View Source Location PART 3 - THE THIRD AXIS CHAPTER 15: DEPLOY NOW, PRAY LATER Can We Automate Uncertainty? Source title Scaling Laws for Reward Model Overoptimization Source authors Leo Gao, John Schulman, Jacob Hilton Source year 2023 Confidence high Importance high
Format Online paper / PDF "Human preference judgments and preference models can sometimes favour agreeable but incorrect responses over truthful ones." View Source Location PART 3 - THE THIRD AXIS CHAPTER 15: DEPLOY NOW, PRAY LATER Can We Automate Uncertainty? Source title Towards Understanding Sycophancy in Language Models Source authors Mrinank Sharma et al. Source year 2023 Confidence high Importance high
Format Video / YouTube Embed "The generator-discriminator gap helps explain why evaluating candidate outputs can be easier than producing a good answer from scratch in some tasks." Location PART 3 - THE THIRD AXIS CHAPTER 15: DEPLOY NOW, PRAY LATER Can We Automate Uncertainty? Source title Deep Dive into LLMs like ChatGPT Source authors Andrej Karpathy Source year 2025 Confidence mid Importance mid
Format Web Page "RLHF can improve instruction following and truthfulness while still relying on imperfect learned signals of human preference." View Source Location PART 3 - THE THIRD AXIS CHAPTER 15: DEPLOY NOW, PRAY LATER Can We Automate Uncertainty? Source title Aligning language models to follow instructions Source authors OpenAI Source year 2022 Confidence high Importance mid
Format Online paper / PDF "In standard RLHF, a separate reward model learns human preferences from ranked outputs and supplies an optimisation signal to the language model." View Source Location PART 3 - THE THIRD AXIS CHAPTER 15: DEPLOY NOW, PRAY LATER Can We Automate Uncertainty? Source title Training language models to follow instructions with human feedback Source authors Long Ouyang et al. Source year 2022 Confidence high Importance high
