Mechanistic tracing shows text suppresses but does not erase audio representations in late layers of Audio LLMs; back-patching reduces text dominance.
arXiv preprint arXiv:2502.17516 , year=
11 Pith papers cite this work. Polarity classification is still indexing.
11
Pith papers citing it
citation-role summary
background 1
citation-polarity summary
roles
background 1polarities
background 1representative citing papers
AI weather models may simulate the atmosphere via particle positions in latent space whose updates follow gradient flow on a learned free energy functional rather than conventional physical equations.
V-SEAM combines concept-level visual semantic editing with attention head modulation to identify positive and negative contributors across object, attribute, and relationship levels, then uses this to improve VLM performance on VQA benchmarks.
Omni-modal LLMs exhibit visual preference that emerges in mid-to-late layers, enabling hallucination detection without task-specific training.
Reasoning in large output spaces proceeds via shortlisting then fine-grained reasoning; this characterization enables a mechanistic distillation strategy that outperforms standard distillation.
Causal mediation analysis on SpiritLM reveals discrepancies in factual recall between text-to-text and speech-to-text paths, indicating only partial carry-over of mechanisms from text to speech modality.
SimReg regularization accelerates LLM pretraining convergence by over 30% and raises average zero-shot performance by over 1% across benchmarks.
HONES ranks feed-forward neurons by their causal contributions from task-relevant attention heads and uses lightweight scaling to steer performance on multiple vision-language tasks.
Large Sensor Models trained on large-scale multimodal wearable data can provide a scalable, general framework for wearable AI by learning transferable representations across modalities and tasks.
The survey organizes mechanistic interpretability techniques into a Locate-Steer-Improve framework to enable actionable improvements in LLM alignment, capability, and efficiency.
A survey categorizing LLM-powered agent systems into software-based, physical, and hybrid types, covering industrial applications and challenges such as latency and security.
citing papers explorer
-
Who Wins the Conflict? Mechanistic Interpretability of Text Bias in Audio LLMs
Mechanistic tracing shows text suppresses but does not erase audio representations in late layers of Audio LLMs; back-patching reduces text dominance.
-
The physics of AI weather models
AI weather models may simulate the atmosphere via particle positions in latent space whose updates follow gradient flow on a learned free energy functional rather than conventional physical equations.
-
V-SEAM: Visual Semantic Editing and Attention Modulating for Causal Interpretability of Vision-Language Models
V-SEAM combines concept-level visual semantic editing with attention head modulation to identify positive and negative contributors across object, attribute, and relationship levels, then uses this to improve VLM performance on VQA benchmarks.
-
Beyond Text-Dominance: Understanding Modality Preference of Omni-modal Large Language Models
Omni-modal LLMs exhibit visual preference that emerges in mid-to-late layers, enabling hallucination detection without task-specific training.
-
Characterize Then Distill: Mechanistic Reasoning in Large Output Spaces
Reasoning in large output spaces proceeds via shortlisting then fine-grained reasoning; this characterization enables a mechanistic distillation strategy that outperforms standard distillation.
-
Do Factual Recall Mechanisms Carry over from Text to Speech in Multimodal Language Models?
Causal mediation analysis on SpiritLM reveals discrepancies in factual recall between text-to-text and speech-to-text paths, indicating only partial carry-over of mechanisms from text to speech modality.
-
SimReg: Achieving Higher Performance in the Pretraining via Embedding Similarity Regularization
SimReg regularization accelerates LLM pretraining convergence by over 30% and raises average zero-shot performance by over 1% across benchmarks.
-
From Heads to Neurons: Causal Attribution and Steering in Multi-Task Vision-Language Models
HONES ranks feed-forward neurons by their causal contributions from task-relevant attention heads and uses lightweight scaling to steer performance on multiple vision-language tasks.
-
Wearable AI in the Era of Large Sensor Models
Large Sensor Models trained on large-scale multimodal wearable data can provide a scalable, general framework for wearable AI by learning transferable representations across modalities and tasks.
-
Locate, Steer, and Improve: A Practical Survey of Actionable Mechanistic Interpretability in Large Language Models
The survey organizes mechanistic interpretability techniques into a Locate-Steer-Improve framework to enable actionable improvements in LLM alignment, capability, and efficiency.
-
LLM-Powered AI Agent Systems and Their Applications in Industry
A survey categorizing LLM-powered agent systems into software-based, physical, and hybrid types, covering industrial applications and challenges such as latency and security.