Beyond the greenwash: do LLMs' ecologically sound decisions hold up under pressure?

arXiv:2610.16043 · cs.CL · Submitted 2026-10-12 · Read on arXiv

cs.CL

Submitted: 2026-10-12

Updated: 2026-10-12

Comments: 21 pages, 6 figures, 6 tables. Model fixtures released at https://huggingface.co/Bioaligned. Companion paper on corpus construction and cross-family installation in preparation

Code: https://github.com/Bioaligned/coupled_welfare_project

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Sources

QLoRA: Efficient Finetuning of Quantized LLMs
Alignment faking in large language models
Alignment Revisited: Are Large Language Models Consistent in Stated and Revealed Preferences?
Measuring Massive Multitask Language Understanding
Designing and Interpreting Probes with Control Tasks
LoRA: Low-Rank Adaptation of Large Language Models
Sleeper Agents: Training Deceptive LLMs that Persist Through Safety Training
Measurement and Fairness
Dynabench: Rethinking Benchmarking in NLP
Bioalignment: Measuring and Improving LLM Disposition Toward Biological Systems for AI Safety
Training language models to follow instructions with human feedback
Do the Rewards Justify the Means? Measuring Trade-Offs Between Rewards and Ethical Behavior in the MACHIAVELLI Benchmark
Discovering Language Model Behaviors with Model-Written Evaluations
Fine-tuning Aligned Language Models Compromises Safety, Even When Users Do Not Intend To!
Safety Alignment Should Be Made More Than Just a Few Tokens Deep
NLP Evaluation in trouble: On the Need to Measure LLM Data Contamination for each Benchmark
Evaluating the Moral Beliefs Encoded in LLMs
Quantifying Language Models' Sensitivity to Spurious Features in Prompt Design or: How I learned to start worrying about prompt formatting
Towards Understanding Sycophancy in Language Models
AI Sandbagging: Language Models can Strategically Underperform on Evaluations

Sources

Related papers