IC-1586In LLaMA-7B, steering all attention heads degrades JSON format accuracy below zero-shot, while steering a subset of 50-100 heads selected via multi-task profiling raises it to 96.64; performance varies dramatically across the 32 layers and individual heads

Qingru Zhang, Chandan Singh, Liyuan Liu, Xiaodong Liu, Bin Yu, Jianfeng Gao, Tuo Zhao

SourceTell Your Model Where to Attend: Post-hoc Attention Steering for LLMs

The paper applies attention reweighting to different subsets of LLaMA-7B's heads and measures the effect on JSON format accuracy. Steering all heads simultaneously drops performance below the zero-shot baseline, indicating that the heads serve distinct and non-redundant functions. Steering an entire layer shows large variation across layers 1-32, and steering individual heads within a layer shows a wide spread (violin plot). Only the heads selected by the multi-task profiling algorithm (intersection of top-k across 4 tasks) yield consistent improvement. The same pattern holds on the pronouns changing, Bias in Bios, and Counterfact tasks.

Evidence
interventional
Key metric
LLaMA-7B JSON format f.acc: zero-shot 60.00, steer all heads below baseline, steer selected heads (k=400-500, |H|=53-86) 96.64; layer-by-layer and head-by-head variation shown in Figure 2
Caveat
The head selection is task-dependent (multi-task setting uses the evaluation task in profiling); the task-agnostic setting (excluding the eval task) achieves 88.16 f.acc, lower than multi-task. The finding is specific to the JSON formatting task in the main figure, though the pattern generalizes to other tasks in the appendix.
Model
LLaMA Llama 7B
Concepts
Depth-dependent structure
Datasets
Bias in Bios [eval], CounterFact / Counterfact dataset [eval]
Related findings
IC-1584, IC-1585
Extraction
automatic-extraction