The authors first examined neuron activations across subjects and transformer layers.
Query-adaptive neuron steering improves frozen language models across benchmarks
Q-NeuS uses each query’s internal activation pattern to select and scale neuron interventions without updating the underlying model.
Chinese Tech
Yanxi Pan · Jingwen Wu · Yan Ma · Xiang Chen · Xinyi Zeng · Xuetong Wang · +5 more
Northeastern University · Washington University in St. Louis · Nanjing University of Aeronautics and Astronautics · Tsinghua University · 360 Security Capability Center
Research Digest··2 min read
Pan and colleagues test whether an LLM’s response to a query can itself identify useful internal interventions.
Why this paper
From Tencent and 6 others
In one line
Q-NeuS adaptively modulates neurons per query to boost LLM performance without retraining or task labels.
What we could check
- ·No code link found
- ·No weights link found
- ·No dataset link found
- ·No compute details found
- ·No stated limitations found
- ·No benchmark numbers found
Observed from the paper text and links we have. Absence here means we did not find it, not that it does not exist.
§