ImProver is an LLM agent using Chain-of-States, error-correction, and retrieval to rewrite Lean proofs for arbitrary user-defined optimization criteria like shortness and readability.
hub Mixed citations
The use of mmr, diversity-based reranking for reordering documents and producing summaries
Mixed citation behavior. Most common role is background (60%).
18
Pith papers citing it
2,144
external citations · external index
Background
60% of classified citations
hub tools
citation-role summary
background 3
method 2
citation-polarity summary
representative citing papers
In multimodal KB-VQA, gold evidence at the first prompt slot beats gold at the last by 16–26 points, flipping the classic U-shaped lost-in-the-middle pattern into primacy bias.
Game-theoretic analysis of diversification in competitive search reveals a diversity-stability tradeoff, with a new method to guarantee corpus equilibrium.
NovelAPIBench is a dynamic benchmark that decomposes novel-API knowledge into bundles, generates executable tasks, and shows usage examples as the strongest signal while retrieval and parametric adaptation play complementary roles.
PRISM benchmark finds LLMs match or exceed humans on isolated review dimensions like novelty verification but none achieve the balanced performance of human reviewers across depth, flaw prioritization, and constructiveness.
RADAGAS-GPT4o achieves a 22.73% bypass rate against 10 WAFs, succeeding more against AI/ML-based firewalls than rule-based ones.
An LLM-based topic modeling method with a custom evaluation framework improves topic interpretability, specificity, and polarity consistency over prior approaches when linking corporate review text to external outcomes such as employee morale.
A question-adaptive greedy frame selector combines SigLIP relevance and DINOv2 coverage under a submodular objective with a text classifier routing to preset trade-offs, yielding accuracy gains on MLVU especially at low frame budgets.
SIFT precomputes selective attention indices via local and cross-attention invariance to speed RAG prefill 1.71x while keeping accuracy within 1% of full recompute, storing only bit vectors 24,000x smaller than KV tensors.
MMTM improves topic coherence and temporal stability in long-form video by tri-modal similarity-gated fusion of speech, audio, and visual embeddings with BERTopic, shown on German and English news datasets with released code and corpus.
RCD balances relevance, coverage, and diversity in a knapsack-constrained selection framework, with experiments showing that selector choice and budget level determine optimal unitization strategies on clinical datasets.
Scientometric review of 15 years of Augmented Human conference papers shows bimodal submission peaks in 2015 and 2025, dominant topics in haptics and wearables, and an active Japanese community alongside definitional scope issues.
A tuning-free LLM pipeline that annotates user histories with inferred motives and uses a reflection loop to correct search queries beats ID-based recommenders on a sparse industrial risk dataset.
DNR is an adversarial denoising neural reranker that extends score error minimization with three objectives to denoise retriever scores and align them with user feedback in two-stage recommender systems.
Applies quantum theory to model incompatibility and interference between relevance dimensions via a cognitive analogue of a quantum physics experiment in a user study.
KITE is a RAG-based tutoring system delivering intent-aware Socratic feedback from course content that improves accuracy of simulated student responses on algorithm tracing and procedural questions.
Argues for a denoising-first paradigm in LLM-oriented information retrieval, framing challenges via a four-stage progression and providing a taxonomy of signal-to-noise optimization techniques across the pipeline.
LaBSE embeddings plus curriculum learning improve macro F1 by up to 0.2 for multilingual multicultural polarization detection in low-resource languages.
citing papers explorer
-
ImProver: Agent-Based Automated Proof Optimization
ImProver is an LLM agent using Chain-of-States, error-correction, and retrieval to rewrite Lean proofs for arbitrary user-defined optimization criteria like shortness and readability.
-
Lost at the End: Primacy Bias in Multimodal Retrieval-Augmented Question Answering
In multimodal KB-VQA, gold evidence at the first prompt slot beats gold at the last by 16–26 points, flipping the classic U-shaped lost-in-the-middle pattern into primacy bias.
-
Stability in Competitive Search with Results Diversification
Game-theoretic analysis of diversification in competitive search reveals a diversity-stability tradeoff, with a new method to guarantee corpus equilibrium.
-
Diagnosing Knowledge Gaps in LLM Tool Use: An Agentic Benchmark for Novel API Acquisition
NovelAPIBench is a dynamic benchmark that decomposes novel-API knowledge into bundles, generates executable tasks, and shows usage examples as the strongest signal while retrieval and parametric adaptation play complementary roles.
-
PRISM: A Multi-Dimensional Benchmark for Evaluating LLM Peer Reviewers
PRISM benchmark finds LLMs match or exceed humans on isolated review dimensions like novelty verification but none achieve the balanced performance of human reviewers across depth, flaw prioritization, and constructiveness.
-
Adversarial SQL Injection Generation with LLM-Based Architectures
RADAGAS-GPT4o achieves a 22.73% bypass rate against 10 WAFs, succeeding more against AI/ML-based firewalls than rule-based ones.
-
Proposing Topic Models and Evaluation Frameworks for Analyzing Associations with External Outcomes: An Application to Leadership Analysis Using Large-Scale Corporate Review Data
An LLM-based topic modeling method with a custom evaluation framework improves topic interpretability, specificity, and polarity consistency over prior approaches when linking corporate review text to external outcomes such as employee morale.
-
Adaptive Greedy Frame Selection for Long Video Understanding
A question-adaptive greedy frame selector combines SigLIP relevance and DINOv2 coverage under a submodular objective with a text classifier routing to preset trade-offs, yielding accuracy gains on MLVU especially at low frame budgets.
-
SIFT: Selective-Index For Fast Compute of RAG Prefill by Exploiting Attention Invariance
SIFT precomputes selective attention indices via local and cross-attention invariance to speed RAG prefill 1.71x while keeping accuracy within 1% of full recompute, storing only bit vectors 24,000x smaller than KV tensors.
-
MMTM: Tri-Modal Topic Modeling for Long-Form Video via Similarity-Gated Fusion
MMTM improves topic coherence and temporal stability in long-form video by tri-modal similarity-gated fusion of speech, audio, and visual embeddings with BERTopic, shown on German and English news datasets with released code and corpus.
-
Budget-Aware Routing for Long Clinical Text
RCD balances relevance, coverage, and diversity in a knapsack-constrained selection framework, with experiments showing that selector choice and budget level determine optimal unitization strategies on clinical datasets.
-
15 Years of Augmented Human(s) Research: Where Do We Stand?
Scientometric review of 15 years of Augmented Human conference papers shows bimodal submission peaks in 2015 and 2025, dominant topics in haptics and wearables, and an active Japanese community alongside definitional scope issues.
-
LLMAR: A Tuning-Free Recommendation Framework for Sparse and Text-Rich Industrial Domains
A tuning-free LLM pipeline that annotates user histories with inferred motives and uses a reflection loop to correct search queries beats ID-based recommenders on a sparse industrial risk dataset.
-
Denoising Neural Reranker for Recommender Systems
DNR is an adversarial denoising neural reranker that extends score error minimization with three objectives to denoise retriever scores and align them with user feedback in two-stage recommender systems.
-
Modelling Dynamic Interactions between Relevance Dimensions
Applies quantum theory to model incompatibility and interference between relevance dimensions via a cognitive analogue of a quantum physics experiment in a user study.
-
Retrieval-Augmented Tutoring for Algorithm Tracing and Problem-Solving in AI Education
KITE is a RAG-based tutoring system delivering intent-aware Socratic feedback from course content that improves accuracy of simulated student responses on algorithm tracing and procedural questions.
-
LLM-Oriented Information Retrieval: A Denoising-First Perspective
Argues for a denoising-first paradigm in LLM-oriented information retrieval, framing challenges via a four-stage progression and providing a taxonomy of signal-to-noise optimization techniques across the pipeline.
-
Leveraging LaBSE with Progressive Curriculum Learning for Multicultural Polarization
LaBSE embeddings plus curriculum learning improve macro F1 by up to 0.2 for multilingual multicultural polarization detection in low-resource languages.