Heads of Security attention gates improve prompt-injection robustness
Authors report RL-Hammer attack success of 10% on HoS versus 97% on GPT-OSS-20B and 85% on SecAlign.
52
AI model
Llama-3.3-70B-Instruct
- Mentions · 7d0
- Mentions · 30d1
- All mentions1
- First seenOct 27, 2026
- Last mentioned15d ago
Timeline
1 articles · newest firstAppears with
- GPT-OSS-20Bai model · 1
Entities are extracted by the model from each article. First seen , last mentioned . Watching keeps it in this browser only (no account); the watchlist page and dashboard alerts use it.