Jun 13, 2026 · Artificial Intelligence

Why PEFT Evaluation Must Go Beyond Downstream Scores: Quantifying General Capability Loss

The PEFT‑Arena benchmark reframes parameter‑efficient fine‑tuning evaluation as a stability‑plasticity trade‑off, measuring both downstream task gains and the preservation of pretrained general abilities through dual‑axis metrics, weight‑space and activation‑space analyses, and path‑wise diagnostics.

Activation GeometryInterpolation AnalysisModel Forgetting

0 likes · 12 min read

Why PEFT Evaluation Must Go Beyond Downstream Scores: Quantifying General Capability Loss

ZhongAn Tech Team

Dec 22, 2024 · Industry Insights

What’s Driving the AI Boom? New Models, Data Limits, and the Rise of Forgetting

This issue reviews the latest AI breakthroughs—including OpenAI’s O3 and o1 models, pricing cuts, new features in ChatGPT, product launches like Pika 2.0 and Gemini 2.0, a heated debate on pre‑training data bottlenecks sparked by Ilya Sutskever, a novel black‑box forgetting method, and DeepMind’s Genie 2 3D world generator—highlighting how industry dynamics and research directions are reshaping the field.

3D generationAIIndustry Trends

0 likes · 12 min read

What’s Driving the AI Boom? New Models, Data Limits, and the Rise of Forgetting