Manifold steering along activation geometry induces behavioral trajectories matching the natural manifold of outputs, while linear steering produces off-manifold unnatural behaviors.
Do Androids Know They’re Only Dreaming of Electric Sheep?
9 Pith papers cite this work, alongside 3 external citations. Polarity classification is still indexing.
9
Pith papers citing it
3
external citations · Crossref
years
2026 9representative citing papers
A learned weighted sum of intermediate-layer activations compresses an instruction prompt into a single patch vector that, injected at an early layer, recovers task accuracy within ~2% of the full prompt.
Steering LLM residual streams with random sparse vectors creates detectable self-recognition fingerprints that enable over 98% accurate attribution of generated text to specific models without degrading output quality.
RISC reformulates self-consistency answer selection as a ranking task solved by a lightweight LambdaRank model with five hand-designed features, yielding better accuracy-efficiency trade-offs than majority voting on QA benchmarks.
TriLens detects hallucinations via per-layer entropy trajectories of logit-lens readouts from three internal modules across LLMs and QA benchmarks.
A factorized study finds raw hidden states and attention features hard to beat in-domain for LLM uncertainty probes, but structured compressed features are more robust under distribution shift, with pretrained probes transferring to open-ended generation.
Disentangling input ambiguity from uncertainty quantification improves error prediction for LLMs on QA tasks, yielding over 10 PRR point gains across models and datasets.
KL divergence of attention heads from uniform distribution predicts LLM answer correctness across datasets and model families.
citing papers explorer
-
Manifold Steering Reveals the Shared Geometry of Neural Network Representation and Behavior
Manifold steering along activation geometry induces behavioral trajectories matching the natural manifold of outputs, while linear steering produces off-manifold unnatural behaviors.
-
Prompt Compression via Activation Aggregation
A learned weighted sum of intermediate-layer activations compresses an instruction prompt into a single patch vector that, injected at an early layer, recovers task accuracy within ~2% of the full prompt.
-
LLM Self-Recognition: Steering and Retrieving Activation Signatures
Steering LLM residual streams with random sparse vectors creates detectable self-recognition fingerprints that enable over 98% accurate attribution of generated text to specific models without degrading output quality.
-
Boosting Self-Consistency with Ranking
RISC reformulates self-consistency answer selection as a ranking task solved by a lightweight LambdaRank model with five hand-designed features, yielding better accuracy-efficiency trade-offs than majority voting on QA benchmarks.
-
TriLens: Per-Layer Logit-Lens Entropy for White-Box Hallucination Detection
TriLens detects hallucinations via per-layer entropy trajectories of logit-lens readouts from three internal modules across LLMs and QA benchmarks.
-
From Signals to Transfer: A Factorised Study of Probe-Based Uncertainty Estimation in Large Language Models
A factorized study finds raw hidden states and attention features hard to beat in-domain for LLM uncertainty probes, but structured compressed features are more robust under distribution shift, with pretrained probes transferring to open-ended generation.
-
The Role of Ambiguity in Error Prediction via Uncertainty Quantification
Disentangling input ambiguity from uncertainty quantification improves error prediction for LLMs on QA tasks, yielding over 10 PRR point gains across models and datasets.
-
Detecting Hallucinations in Large Language Models via Internal Attention Divergence Signals
KL divergence of attention heads from uniform distribution predicts LLM answer correctness across datasets and model families.
- Process Supervision of Confidence Margin for Calibrated LLM Reasoning