The paper investigates how activation steering—adding a vector to a large language model’s residual-stream activations—alters discourse-level aspects of generated text, such as formality, evidence usage, and ambiguity. It demonstrates that these unintended changes can be largely mitigated by applying a second, complementary steering vector that targets the affected discourse dimension, restoring about 78% of the original discourse characteristics while maintaining the intended behavior in roughly 93% of outputs.
arXiv:2607. 18570v1 Announce Type: cross Abstract: Discourse relations provide document structure, critical to language understanding and enabling language model performance and ethicality.
arXiv:2606.08682v2 Announce Type: replace Abstract: Activation steering (AS) has emerged as a popular inference-time technique for modulating the behavior of large language models (LLMs). AS modulate...
arXiv:2606. 27598v1 Announce Type: cross Abstract: Ultra-fine entity typing (UFET) assigns highly specific types to entity mentions, but current approaches struggle with types in the long tail.
arXiv:2605. 09159v2 Announce Type: replace Abstract: Recent work shows that large language models (LLMs) encode behavioral traits ("personas") as linear directions in activation space, often called "persona vectors".
arXiv:2606. 08682v1 Announce Type: cross Abstract: Activation steering has emerged as a popular inference-time technique for modulating the behavior of large language models (LLMs).
arXiv:2610.28092v1 Announce Type: cross Abstract: Persuasive writing exploits cognitive biases, recruiting a reader's mental shortcuts to shape belief, often without the reader noticing. We call the...
The paper investigates how different prompt adaptation techniques—discrete prompt optimization, prompt tuning, and prefix tuning—affect the internal processing of dense and mixture-of-experts language models. It finds that the adaptation location determines how task information reaches the output: text prompts mainly influence response construction, input-level prompts modify content representations, and layer-wise prefixes exert direct control via attention. In MoE models, these differences also lead to distinct expert‑routing patterns, while generic representation‑level metrics fail to reliably distinguish effective adaptations from placebo effects.
arXiv:2607. 25270v1 Announce Type: cross Abstract: Activation steering controls language models by adding vectors or features to hidden states at inference time, but the upstream source of these steering signals is often treated as a secondary detail.
arXiv:2608. 12334v1 Announce Type: cross Abstract: Despite the impressive multilingual capabilities of Large Language Models, the latent dynamics dictating language selection remain poorly understood.
arXiv:2507. 18043v2 Announce Type: replace-cross Abstract: Inference-time steering methods offer a lightweight alternative to fine-tuning large language models (LLMs) and vision-language models (VLMs) by modifying internal activations at test time without updating model weights.
Large language models (LLMs) are widely used to assist writing, but this study shows they alter both tone and meaning of human text. A user study found that heavy LLM use increased neutral essays by nearly 70% and made writers feel less creative and less in their voice. Even when prompted to make only grammar edits, LLMs changed the semantic content of essays and produced AI-generated scientific reviews that were less focused on clarity and significance and scored higher on average.
The paper investigates how adjectival modifiers affect the semantic plausibility of events, using the Adept benchmark of 16,000 English sentence pairs that differ by a single adjective. Experiments show that sentence transformers, despite being conceptually suited to the task, underperform compared to models like RoBERTa. The authors provide an error analysis and discuss the implications of their findings for future work on balancing training and test data.