Introduces CoSee auditing framework and identifies Noise Reinforcement and Policy Collapse as dominant failure modes when weak 4B-8B models use shared state for multi-page visual QA.
emnlp-main.243/
4 Pith papers cite this work. Polarity classification is still indexing.
4
Pith papers citing it
years
2026 4verdicts
UNVERDICTED 4representative citing papers
EPIC constructs a preference-aligned index for on-device RAG that reduces memory 2404x, cuts retrieval latency 32x, and raises preference-following accuracy 18.79pp across four benchmarks while fitting under 1MB on real devices.
Language models acquire skills during pretraining in a consistent compositional order, with composite tasks emerging after their components and trajectories predictable from model representations.
PRISM is a gauge-invariant DP mechanism for LoRA that avoids bilinear noise amplification via tangent-space sampling, supplies a closed-form noise characterization on Z, and includes a DP-aware adaptive update rule.
citing papers explorer
-
Diagnosing Failure Modes of Shared-State Collaboration in Resource-Constrained Visual Agents
Introduces CoSee auditing framework and identifies Noise Reinforcement and Policy Collapse as dominant failure modes when weak 4B-8B models use shared state for multi-page visual QA.
-
From Volume to Value: Preference-Aligned Memory Construction for On-Device RAG
EPIC constructs a preference-aligned index for on-device RAG that reduces memory 2404x, cuts retrieval latency 32x, and raises preference-following accuracy 18.79pp across four benchmarks while fitting under 1MB on real devices.
-
What do Language Models Learn and When? The Implicit Curriculum Hypothesis
Language models acquire skills during pretraining in a consistent compositional order, with composite tasks emerging after their components and trajectories predictable from model representations.
-
PRISM: Gauge-Invariant Tangent-Space Differentially Private LoRA
PRISM is a gauge-invariant DP mechanism for LoRA that avoids bilinear noise amplification via tangent-space sampling, supplies a closed-form noise characterization on Z, and includes a DP-aware adaptive update rule.