archive
Every paper Pith has read. Search by title, abstract, or pith.
1286 papers in cs.IR · page 10
-
Normalization plus graph alignment removes LLM recommender training barrier
Break the Optimization Barrier of LLM-Enhanced Recommenders: A Theoretical Analysis and Practical Framework
-
Hybrid detector finds 893k eliminable duplicate BDD steps
Reducing Maintenance Burden in Behaviour-Driven Development: A Paraphrase-Robust Duplicate-Step Detector with a 1.1M-Step Open Benchmark
-
HaS cuts RAG retrieval latency by up to 37 percent
HaS: Accelerating RAG through Homology-Aware Speculative Retrieval
-
Discrete tokens turn user multimodal interactions into personalized text and images
Discrete Preference Learning for Personalized Multimodal Generation
-
Semantic recall fixes vector search evaluation by ignoring irrelevant neighbors
Semantic Recall for Vector Search
-
Unified library gathers over twenty metrics for SPARQL query evaluation
T2S-Metrics: Unified Library for Evaluating SPARQL Queries Generated From Natural Language
-
R³AG picks retrievers by both relevance and answer-generation utility
R$^3$AG: Retriever Routing for Retrieval-Augmented Generation
-
Model learns to judge when to retrieve external knowledge for entity recognition
SAKE: Self-aware Knowledge Exploitation-Exploration for Grounded Multimodal Named Entity Recognition
-
MLLM attributes sharpen product retrieval in e-commerce
AFMRL: Attribute-Enhanced Fine-Grained Multi-Modal Representation Learning in E-commerce
-
Knowledge graphs learn their own forgetting rates from data
Not All Memories Age the Same: Autodiscovery of Adaptive Decay in Knowledge Graphs
-
LLM agents could turn hidden profiles into governable personalization
From Hidden Profiles to Governable Personalization: Recommender Systems in the Age of LLM Agents
-
Five agents plus skill hub automate recsys config tuning
AgenticRecTune: Multi-Agent with Self-Evolving Skillhub for Recommendation System Optimization
-
LLM agents automate recsys configuration optimization
AgenticRecTune: Multi-Agent with Self-Evolving Skillhub for Recommendation System Optimization
-
MetaRAG reproduces with relative gains but lower absolute scores
A Reproducibility Study of Metacognitive Retrieval-Augmented Generation
-
4B agent beats larger models on research tasks with 10K open data
DR-Venus: Towards Frontier Edge-Scale Deep Research Agents with Only 10K Open Data
-
ECLASS boosts dense retrieval to 94% hit rate on component queries
ECLASS-Augmented Semantic Product Search for Electronic Components
-
Benchmark shows CE trade-offs for recommenders vary by method and format
From Top-1 to Top-K: A Reproducibility Study and Benchmarking of Counterfactual Explanations for Recommender Systems
-
AI conference reviews lengthened and standardized after LLMs
Impact of large language models on peer review opinions from a fine-grained perspective: Evidence from top conference proceedings in AI
-
ColBERT embeddings aligned to clinical concepts for easier debugging
Diagnosable ColBERT: Debugging Late-Interaction Retrieval Models Using a Learned Latent Space as Reference
-
CTR models train with loops but infer in one pass
LoopCTR: Unlocking the Loop Scaling Power for Click-Through Rate Prediction
-
Structure matching in RAG improves factual answers by up to 50 points
Structure Guided Retrieval-Augmented Generation for Factual Queries
-
Highlights and abstracts together improve keyword extraction
Enhancing Unsupervised Keyword Extraction in Academic Papers through Integrating Highlights with Abstract
-
Masking and retriever ensembles correct facts 14% better
Mask-to-Correct$^+$: Leveraging Retriever Diversity for Masking-guided Faithful Fact Correction
-
Semantic code transitions improve next-item predictions
CAST: Modeling Semantic-Level Transitions for Complementary-Aware Sequential Recommendation
-
IndiaFinBench gives first public test of LLMs on Indian financial rules
IndiaFinBench: An Evaluation Benchmark for Large Language Model Performance on Indian Financial Regulatory Text
-
CS3 lifts two-tower ad revenue 8.36% at ms latency
CS3: Efficient Online Capability Synergy for Two-Tower Recommendation
-
Query-aware diffusion recovers relevant graph subgraphs with guarantees
Query-Aware Flow Diffusion for Graph-Based RAG with Retrieval Guarantees
-
GraphRAG-IRL fuses graph IRL with LLM re-ranking for recommendations
GraphRAG-IRL: Personalized Recommendation with Graph-Grounded Inverse Reinforcement Learning and LLM Re-ranking
-
Feature optimization raises AI citation rates without quality loss
Think Before Writing: Feature-Level Multi-Objective Optimization for Generative Citation Visibility
-
Retriever performance drops sharply on redundant corpora
RARE: Redundancy-Aware Retrieval Evaluation Framework for High-Similarity Corpora
-
Adapter with three MoE modules improves temporal graph event forecasts
STK-Adapter: Incorporating Evolving Graph and Event Chain for Temporal Knowledge Graph Extrapolation
-
Most users rank LLMs differently from aggregate leaderboards
Personalized Benchmarking: Evaluating LLMs by Individual Preferences
-
Dual-view training raises instruction-following retrieval 45%
Dual-View Training for Instruction-Following Information Retrieval
-
New dataset challenges AI on global math contests
MathNet: a Global Multimodal Benchmark for Mathematical Reasoning and Retrieval
-
Text embeddings beat page images for science paper search
Document-as-Image Representations Fall Short for Scientific Retrieval
-
Gaussian limits bound TF-IDF errors from lost tokens
Context-Aware Search and Retrieval Under Token Erasure
-
ArbGraph arbitration cuts hallucinations in long-form RAG
ArbGraph: Conflict-Aware Evidence Arbitration for Reliable Long-Form Retrieval-Augmented Generation
-
Balanced clustering cuts recommender embeddings by 75%
Balanced Co-Clustering of Users and Items for Embedding Table Compression in Recommender Systems
-
Trie guidance lets T5 and BART beat larger models on document queries
DocQAC: Adaptive Trie-Guided Decoding for Effective In-Document Query Auto-Completion
-
Full-context LLM judging outperforms isolated checks for multi-hop RAG
Evaluating Multi-Hop Reasoning in RAG Systems: A Comparison of LLM-Based Retriever Evaluation Strategies
-
Multi-LLM filtering improves sequential recommendations without text
Multi-LLM Token Filtering and Routing for Sequential Recommendation
-
Modular control fixes LLM final-layer weakness for recommendations
Modular Representation Compression: Adapting LLMs for Efficient and Effective Recommendations
-
Human-AI collaboration gaps trace to mismatched grounding capacity
The Collaboration Gap in Human-AI Work
-
Memory wins for precise finance math
Architecture Matters More Than Scale: A Comparative Study of Retrieval and Memory Augmentation for Financial QA Under SME Compute Constraints
-
Memory wins precise finance math
Architecture Matters More Than Scale: A Comparative Study of Retrieval and Memory Augmentation for Financial QA Under SME Compute Constraints
-
Gaussian process beats passive LLM reranking for passage retrieval
Bayesian Active Learning with Gaussian Processes Guided by LLM Relevance Scoring for Dense Passage Retrieval
-
RankUp lifts ad GMV up to 4.81% by preserving higher-rank representations
RankUp: Towards High-rank Representations for Large Scale Advertising Recommender Systems
-
RankUp lifts ad recommender GMV up to 4.81% by raising representation rank
RankUp: Towards High-rank Representations for Large Scale Advertising Recommender Systems
-
Semantic bridge enables private cross-domain recs without overlaps
FedCRF: A Federated Cross-domain Recommendation Method with Semantic-driven Deep Knowledge Fusion
-
150k AI papers show must-cite retrieval still unsolved
MasterSet: A Large-Scale Benchmark for Must-Cite Citation Recommendation in the AI/ML Literature