Chain of Evidence introduces a retriever-agnostic visual attribution method for iRAG that reasons over document screenshots with VLMs to output precise bounding boxes, outperforming text baselines on Wiki-CoE and SlideVQA.
Financial analysis: Intelligent financial data analysis system based on llm-rag
5 Pith papers cite this work. Polarity classification is still indexing.
5
Pith papers citing it
citation-role summary
background 2
citation-polarity summary
verdicts
UNVERDICTED 5roles
background 2polarities
background 2representative citing papers
EVGeoQA benchmark and GeoRover framework show LLMs can use tools for sub-tasks in dynamic geo-spatial exploration but struggle with long-range planning, with an emergent ability to improve via historical trajectory summaries.
CoRM-RAG uses a cognitive perturbation protocol to simulate biases and trains an Evidence Critic to retrieve documents that support correct decisions even under adversarial query changes.
SCOUT uses token saliency analysis to detect both standard and contextually-plausible backdoor attacks in language models while maintaining clean accuracy.
LoRA-MINT uses perplexity to perform membership inference on LoRA-fine-tuned LLMs, reporting 0.77-0.92 precision across four models and three datasets while outperforming baselines.
citing papers explorer
-
Chain of Evidence: Pixel-Level Visual Attribution for Iterative Retrieval-Augmented Generation
Chain of Evidence introduces a retriever-agnostic visual attribution method for iRAG that reasons over document screenshots with VLMs to output precise bounding boxes, outperforming text baselines on Wiki-CoE and SlideVQA.
-
EVGeoQA: Benchmarking LLMs on Dynamic, Multi-Objective Geo-Spatial Exploration
EVGeoQA benchmark and GeoRover framework show LLMs can use tools for sub-tasks in dynamic geo-spatial exploration but struggle with long-range planning, with an emergent ability to improve via historical trajectory summaries.
-
Beyond Semantic Relevance: Counterfactual Risk Minimization for Robust Retrieval-Augmented Generation
CoRM-RAG uses a cognitive perturbation protocol to simulate biases and trains an Evidence Critic to retrieve documents that support correct decisions even under adversarial query changes.
-
SCOUT: A Defense Against Data Poisoning Attacks in Fine-Tuned Language Models
SCOUT uses token saliency analysis to detect both standard and contextually-plausible backdoor attacks in language models while maintaining clean accuracy.
-
Auditing Training Data in Domain-adapted LLMs: LoRA-MINT
LoRA-MINT uses perplexity to perform membership inference on LoRA-fine-tuned LLMs, reporting 0.77-0.92 precision across four models and three datasets while outperforming baselines.