QAOD projects away question-aligned directions from answer representations to isolate domain-agnostic factuality signals, enabling efficient hallucination detection with top in-domain AUROC and up to 21% better OOD transfer.
Title resolution pending
9 Pith papers cite this work. Polarity classification is still indexing.
9
Pith papers citing it
citation-role summary
background 1
citation-polarity summary
years
2026 9roles
background 1polarities
unclear 1representative citing papers
Ranking latent-memory fragments by middle-layer attention density and keeping only the top few more than doubles long-term knowledge-retention accuracy (43.0% vs. 17.4%/17.6%) versus MemoryLLM and M+.
Linear probes on frozen LLM hidden states recover an approximate remaining-output-length signal that is decodable at prompt-end, transfers across datasets, and shifts upward at retraction tokens.
Empirical evaluation on the PLUME benchmark shows steering vectors vary widely in trait expressibility, degrade on task transfer, and lose effectiveness when multiple vectors are composed.
Contrastive Logit Steering isolates a linear refusal direction in safety-aligned LLMs, achieving higher jailbreak success than activation steering and enabling bidirectional control without retraining.
COAST applies contrastive conceptors to steer VLA hidden states into task-specific success subspaces, yielding over 20% simulation and 40% real-robot success rate gains across three distinct policies.
Stable personality vectors in LLMs function as intrinsic guardrails, with ablation increasing emergent misalignment above 40% and amplification reducing it below 3%, enabling zero-shot transfer from aligned to corrupted models.
SIREN identifies safety neurons via linear probing on internal LLM layers and combines them with adaptive weighting to detect harm, outperforming prior guard models with 250x fewer parameters.
A geometric 1-form on token embeddings has curvature that couples to semantic world models in language models, as evidenced by clustering on chess board regions and piece importance.
citing papers explorer
-
When Answers Stray from Questions: Hallucination Detection via Question-Answer Orthogonal Decomposition
QAOD projects away question-aligned directions from answer representations to isolate domain-agnostic factuality signals, enabling efficient hallucination detection with top in-domain AUROC and up to 21% better OOD transfer.
-
MemDefrag: Latent Memory Defragmentation for Large Language Models
Ranking latent-memory fragments by middle-layer attention density and keeping only the top few more than doubles long-term knowledge-retention accuracy (43.0% vs. 17.4%/17.6%) versus MemoryLLM and M+.
-
How Much is Left? LLMs Linearly Encode Their Remaining Output Length
Linear probes on frozen LLM hidden states recover an approximate remaining-output-length signal that is decodable at prompt-end, transfers across datasets, and shifts upward at retraction tokens.
-
On the Limits of Steering Vectors for Preference-Aligned Generation
Empirical evaluation on the PLUME benchmark shows steering vectors vary widely in trait expressibility, degrade on task transfer, and lose effectiveness when multiple vectors are composed.
-
The Geometry of Refusal: Linear Instability in Safety-Aligned LLMs
Contrastive Logit Steering isolates a linear refusal direction in safety-aligned LLMs, achieving higher jailbreak success than activation steering and enabling bidirectional control without retraining.
-
Contrastive Conceptor Activation Steering (COAST): Unlocking Vision-Language-Action Models through Hidden States
COAST applies contrastive conceptors to steer VLA hidden states into task-specific success subspaces, yielding over 20% simulation and 40% real-robot success rate gains across three distinct policies.
-
Intrinsic Guardrails: How Semantic Geometry of Personality Interacts with Emergent Misalignment in LLMs
Stable personality vectors in LLMs function as intrinsic guardrails, with ablation increasing emergent misalignment above 40% and amplification reducing it below 3%, enabling zero-shot transfer from aligned to corrupted models.
-
LLM Safety From Within: Detecting Harmful Content with Internal Representations
SIREN identifies safety neurons via linear probing on internal LLM layers and combines them with adaptive weighting to detect harm, outperforming prior guard models with 250x fewer parameters.
-
A geometric relation of the error introduced by sampling a language model's output distribution to its internal state
A geometric 1-form on token embeddings has curvature that couples to semantic world models in language models, as evidenced by clustering on chess board regions and piece importance.