Skip to content
conftrace_
2026 EMNLP EMNLP 2026

The Attentional White Bear: Persistent But Attenuated Concept Recoverability in Transformer Language Models

AbstractInstruction-based suppression is widely used to prevent language models from generating prohibited content, yet it remains unclear whether suppression reduces internal representation or merely suppresses expression. We investigate this question through representational probing, attention analysis, and behavioral semantic leakage experiments across multiple transformer models. We find that prohibited concepts remain highly recoverable from hidden representations under suppression, influence localized changes in attention routing, and measurably shape downstream generations despite successful lexical avoidance. These effects persist across pooling strategies, indirect semantic controls, and multiple model families. Our results expose a fundamental gap between behavioral and representational alignment.

🌉 Interdisciplinary Bridge - Artificial Intelligence and Interdisciplinary and Natural Language Processing

More from EMNLP 2026

All EMNLP 2026 papers →