mlsanityData to 7 Oct 2026

Alignment midtraining for animals

arXiv 2604.13076 ↗First posted 21 Mar 2026cs.CL cs.AIcited by 4

Not on the map. Why?

Abstract (arXiv)

We investigate the robustness of value alignment via midtraining with synthetic documents, using animal compassion as a value that is both important in its own right and orthogonal to existing alignment efforts. To evaluate compassionate reasoning, we develop and publicly release Animal Norms In Moral Assessment (ANIMA), a 26-question evaluation spanning 13 ethical dimensions, publicly available as a dataset and Inspect evaluation. On ANIMA, training with 3000 documents achieves 77% compared to 40% for instruction-tuning approaches, with generalization to human compassion and no degradation in standard safety benchmarks or capabilities. However, subsequent unrelated instruction-tuning degrades the intervention, with the advantage disappearing after 5000 samples. Our exploratory results suggest document-based value interventions may require explicit preservation strategies to remain effective through typical training pipelines.

Key facts. Cited by 4 papers in the map (data through 7 Oct 2026). Builds on 32 of its 54 references that are in the map. First posted on arXiv 21 Mar 2026.

Builds on

32 of 54 references.

References · 32

  1. 2022Training language models to follow instructions with human feedbackcited by 15,712
  2. 2022Scaling Instruction-Finetuned Language Modelscited by 2,997
  3. 2022Constitutional AI: Harmlessness from AI Feedbackcited by 2,950
  4. 2023LIMA: Less Is More for Alignmentcited by 1,080
  5. 2024Alignment faking in large language modelscited by 316
  6. 2025Persona Vectors: Monitoring and Controlling Character Traits in Language Modelscited by 303
  7. 2023The Unlocking Spell on Base LLMs: Rethinking Alignment via In-Context Learningcited by 251
  8. 2023A Pretrainer's Guide to Training Data: Measuring the Effects of Data Age, Domain Coverage, Quality, & Toxicitycited by 180
  9. 2023Linear Representations of Sentiment in Large Language Modelscited by 157
  10. 2023Taken out of context: On measuring situational awareness in LLMscited by 131
  11. 2025Natural Emergent Misalignment from Reward Hacking in Production RLcited by 111
  12. 2025Subliminal Learning: Language models transmit behavioral traits via hidden signals in datacited by 89
20 more
  1. 2025Persona Features Control Emergent Misalignmentcited by 76
  2. 2026The Assistant Axis: Situating and Stabilizing the Default Persona of Language Modelscited by 76
  3. 2025Auditing language models for hidden objectivescited by 68
  4. 2025Poisoning Attacks on LLMs Require a Near-constant Number of Poison Samplescited by 66
  5. 2024Language Models Resist Alignment: Evidence From Data Compressioncited by 50
  6. 2025Weird Generalization and Inductive Backdoors: New Ways to Corrupt LLMscited by 47
  7. 2025On the generalization of language models from in-context learning and finetuning: a controlled studycited by 39
  8. 2025Inoculation Prompting: Instructing LLMs to misbehave at train-time improves test-time alignmentcited by 32
  9. 2024Dissecting Fine-Tuning Unlearning in Large Language Modelscited by 25
  10. 2026Alignment Pretraining: AI Discourse Causes Self-Fulfilling (Mis)alignmentcited by 25
  11. 2025Demystifying Synthetic Data in LLM Pre-training: A Systematic Study of Scaling Laws, Benefits, and Pitfallscited by 24
  12. 2025On Linear Representations and Pretraining Data Frequency in Language Modelscited by 23
  13. 2025Believe It or Not: How Deeply do LLMs Believe Implanted Facts?cited by 19
  14. 2025How to Alleviate Catastrophic Forgetting in LLMs Finetuning? Hierarchical Layer-Wise and Element-Wise Regularizationcited by 13
  15. 2025The Psychogenic Machine: Simulating AI Psychosis, Delusion Reinforcement and Harm Enablement in Large Language Modelscited by 8
  16. 2025Jailbreak-Tuning: Models Efficiently Learn Jailbreak Susceptibilitycited by 6
  17. 2025Made-in China, Thinking in America:U.S. Values Persist in Chinese LLMscited by 5
  18. 2024How far can bias go? Tracing bias from pretraining data to alignmentcited by 4
  19. 2025The Devil in the Details: Emergent Misalignment, Format and Coherence in Open-Weights LLMscited by 4
  20. 2025On the Robustness Tradeoff in Fine-Tuningcited by 3

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2026: 4.

4 papers.

  1. 2026Coercion and Deception in AI-to-AI Management: An Agentic Benchmark of Unprompted Escalationcited by 2
  2. 2026Do LLMs Hold Their Values? MANTA: A Multi-Turn Adversarial Benchmark for Animal Welfare Reasoningcited by 1
  3. 2026Helpfulness Hurts: Domain-Dependent Degradation of Mid-Trained Compassion Values Under Post-Trainingcited by 0
  4. 2026HarvestBench: Measuring Whether LLM Agents Will Pay to Avoid Killing Animalscited by 0