{
  "id": 10070,
  "url": "https://arxiv.org/abs/2606.26102",
  "title": "Helpfulness Hurts: Domain-Dependent Degradation of Mid-Trained Compassion Values Under Post-Training",
  "summary": "arXiv:2606.26102v3 Announce Type: replace-cross Abstract: Standard post-training pipelines apply supervised fine-tuning (SFT) and reinforcement learning (RL) to make language models helpful, but these processes may inadvertently degrade values instilled during pre-training. We investigate whether the domain of post-training data differentially affects the retention of animal compassion values in a Llama 3.1 8B model mid-trained on compassion-oriented synthetic data, using both SFT (helpfulness v",
  "authors": "Jasmine Brazilek, Juliana Seawell",
  "category": "research",
  "topics": "finance-investment",
  "orgs": "meta",
  "regions": null,
  "published_at": "2026-07-14T04:00:00.000Z",
  "fetched_at": "2026-07-14T16:55:59.928Z",
  "source_slug": "arxiv-cscy",
  "source_name": "arXiv cs.CY",
  "source_homepage": "https://arxiv.org/list/cs.CY/recent",
  "ethics_ai_record_url": "https://ethics.ai/record/10070",
  "original_url": "https://arxiv.org/abs/2606.26102",
  "evidence_status": "source-only",
  "attribution": "via ethics.ai"
}