Jasmine Brazilek — LessWrong

Assert, don't describe: how writing style in training data shapes an AI's moral stance

tl;dr The way things are said ("linguistic features") in fine-tuning data affect AI alignment. Some features degrade alignment to the value targeted in the corpus, some have negligible impact, and others bolster it. We recommend you become familiar with these features if your writing may end up in LLM training...

Jun 172

[Linkpost] Community polls on alignment controversies

Planning where we focus at CaML requires forming views on many controversial questions. In many cases, people we've talked to have wildly different perceptions of the balance of opinions, so we thought this would be a great way for us and others to know where we're out of step on...

Jun 178

Document-tuning instills durable animal compassion in LLMs (and generalizes to humans)

Note: This post focuses on the alignment implications. Our EA Forum, focusing on the implications for animal welfare, is here. Jasmine Brazilek & Miles Tidmarsh: Compassion Aligned Machine Learning Preprint, March 2026: Full paper | HuggingFace resources | ANIMA benchmark TL;DR Instruction-tuning and Reinforcement learning are effective in certain specific...

May 2111

Document-tuning instills durable animal compassion in LLMs (and generalizes to humans)

Note: This post focuses on the alignment implications. Our EA Forum, focusing on the implications for animal welfare, is here. Jasmine Brazilek & Miles Tidmarsh: Compassion in Machine Learning Preprint, March 2026: Full paper | HuggingFace resources | Animal Harm Benchmark (AHB) TL;DR Instruction-tuning and RL are effective in-context but...

May 151

Make the future non-human beings deserve ($5k in prizes)

Summary If you think future AI systems should extend moral consideration to non-human sentient beings, there’s something concrete you can do about it right now. It’s free, it takes thirty seconds, and you don’t need to log in. Go to hyperstition.sentientfutures.ai, type a prompt about animal welfare or digital sentience,...

Apr 11

Teaching AI Compassion Through Synthetic Documents

Jasmine Brazilek & Miles Tidmarsh: Compassion Aligned Machine Learning Preprint, March 2026: Full paper (PDF) | HuggingFace resources TL;DR We show that fine-tuning language models on a small number of synthetic midtraining/pretraining-style[1] documents (not instruction-tuning data) can improve their compassionate reasoning about animals. Our document-tuned model scores 77% on a...

Apr 11

Your AI Travel agent would book you a bullfight: benchmarking implicit animal compassion in Agentic AI

By Compassion Aligned Machine Learning (CaML) AI agents are being used to act on our behalf: booking flights, planning holidays, making purchases, but what happens when the most relevant search result is an elephant trekking tour, a dolphin captivity show, or tickets to a bullfight? We built TAC (Travel Agent...

Mar 317