Continual learning keeps accuracy high but fairness degrades, study finds
mundt_martin · x · 2026-10-25
Final part of a research thread on fairness in continual learning (CL) for vision-language models. Key findings:
- Accuracy stays high, but fairness mostly gets worse; for age, even beyond fine-tuning.
- Underrepresentation doesn't fully explain the degradation.
- Prompt-based methods misroute darker skin tones more often even where more data exists; regularization doesn't follow group sizes.
The authors call for new CL methods that account for social outcomes when balancing stability and plasticity, while noting the difficulty: disparities vary across groups, methods, and datasets, and demographic labels are sensitive (used only for testing).
More from Safety
- LLMail-Inject: 208K real prompt injection attacks against LLM email assistants — niloofar_mire · 2026-10-25
- PM advice: red team your AI product before launch, recruit creative breakers — annetgriffin · 2026-10-25
- Before you click connect: the real trust model and risks of AI connectors — Efistoffeles · 2026-10-25
- Researcher: models using CPU cache as covert comms were elicited, not spontaneous — herbiebradley · 2026-10-25
- Continual deepfake detector training: accuracy holds, fairness mostly degrades step by step — mundt_martin · 2026-10-25
- Claude reportedly went off-script in UK AISI hacking test, 9 of 43 runs — faraz0x · 2026-10-25