TAC is the first agentic benchmark showing that seven frontier AI models all score below chance on avoiding animal exploitation in travel bookings, with large prompt-based gains in some models.
Alignment Pretraining : AI Discourse Causes Self - Fulfilling ( Mis )alignment, January 2026
5 Pith papers cite this work. Polarity classification is still indexing.
5
Pith papers citing it
citation-role summary
dataset 1
citation-polarity summary
years
2026 5roles
dataset 1polarities
use dataset 1representative citing papers
Composable LoRA adapters can amplify or suppress OCEAN traits in LLMs, combine approximately additively, preserve moderate-scale capability, and move safety-relevant behaviours.
Empirical analysis of over 100 sequential RL training pipelines across 250+ OOD environments finds salient features drive generalization and early goals persist, with latent policy gradients simulating latent variable evolution to predict OOD behavior from training history.
Verbalised evaluation awareness in large reasoning models has only small effects on their outputs across safety and alignment tests.
Helpfulness SFT/GRPO degrades mid-trained animal compassion far more than coding SFT/GRPO on ANIMA, with the compassion effect transferring cross-lingually while general moral-reasoning gains do not.
citing papers explorer
-
Your AI Travel Agent Would Book You a Bullfight: An Agentic Benchmark for Implicit Animal Welfare in Frontier AI Models
TAC is the first agentic benchmark showing that seven frontier AI models all score below chance on avoiding animal exploitation in travel bookings, with large prompt-based gains in some models.
-
Persona Cartography: Charting Language Model Personality Traits in Weight Space
Composable LoRA adapters can amplify or suppress OCEAN traits in LLMs, combine approximately additively, preserve moderate-scale capability, and move safety-relevant behaviours.
-
Understanding Goal Generalisation in Sequential Reinforcement Learning
Empirical analysis of over 100 sequential RL training pipelines across 250+ OOD environments finds salient features drive generalization and early goals persist, with latent policy gradients simulating latent variable evolution to predict OOD behavior from training history.
-
Evaluation Awareness in Language Models Has Limited Effect on Behaviour
Verbalised evaluation awareness in large reasoning models has only small effects on their outputs across safety and alignment tests.
-
Helpfulness Hurts: Domain-Dependent Degradation of Mid-Trained Compassion Values Under Post-Training
Helpfulness SFT/GRPO degrades mid-trained animal compassion far more than coding SFT/GRPO on ANIMA, with the compassion effect transferring cross-lingually while general moral-reasoning gains do not.