A training-free two-stage token compression framework preserves 92.9% of Qwen2.5-Omni-7B's audio-visual understanding accuracy while using 6.8% of the original multimodal-token FLOPs.
ContextGuard: Structured Self-Auditing for Context Learning in Language Models
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
abstract
Recent benchmarks reveal that despite strong reasoning capabilities, large language models (LLMs) still struggle to faithfully apply complex contextual knowledge. These failures are often not wholesale reasoning collapses: in context-rich tasks, models may follow the central reasoning path while missing peripheral, persistent, or format-sensitive requirements.
citation-role summary
background 1
citation-polarity summary
fields
cs.CV 1years
2026 1verdicts
CONDITIONAL 1roles
background 1polarities
unclear 1representative citing papers
citing papers explorer
-
OmniPack: Unified Token Compression for Efficient Omni-modal Large Language Models
A training-free two-stage token compression framework preserves 92.9% of Qwen2.5-Omni-7B's audio-visual understanding accuracy while using 6.8% of the original multimodal-token FLOPs.