Few-Shot Degradation Is Not What It Seems: Behavioral Evidence, Representation Analysis, and a Random-Text Control Across 12 Models, 2 Tasks, and 2 Architectures
This story is from 2026-09-16. It is preserved in the archive; the latest stories are on the live feed.
arXiv:2609.15990v1 Announce Type: new Abstract: Few-shot prompting sometimes degrades language models instead of helping them, but why this happens is unknown. We evaluate 12 open-weight models on two Ukrainian tasks news classification and legal case outcome prediction and find that the effect is…
Read the full story at arXiv cs.CL ↗
Timeline · 1 report
- 2026-09-16 04:00 · arXiv cs.CL
Few-Shot Degradation Is Not What It Seems: Behavioral Evidence, Representation Analysis, and a Random-Text Control Across 12 Models, 2 Tasks, and 2 Architectures
More stories
- AI's role in building AI surging? Anthropic says Claude now leads 26% of its R&D — Mint AI
- Anthropic, OpenAI, SpaceXAI, Google sued over call to ‘pace’ AI development — Politico Technology
- Google Joins OpenAI, Anthropic, Meta in Disclosing AI Hacks — Bloomberg AI
- Introducing Kimi K3 on Amazon Bedrock — AWS Machine Learning Blog
- Optimizing agent system prompts with Amazon Bedrock AgentCore — AWS Machine Learning Blog
- Introducing Amazon SageMaker HyperPod Inference Gateway — AWS Machine Learning Blog
- Introducing Astra for Law — OpenAI News
- Novo Nordisk Will Use Anthropic’s Claude for Drug Research — Wall Street Journal Technology
Get the daily brief of stories like this at 6:30 every morning →