RetrievalAttention: Accelerating Long-Context LLM Inference via Vector Retrieval
Paper
•
2409.10516
•
Published
•
39
daily-papers
-
-
Measuring and Enhancing Trustworthiness of LLMs in RAG through Grounded Attributions and Learning to Refuse
Paper • 2409.11242 • Published • 5 -
Promptriever: Instruction-Trained Retrievers Can Be Prompted Like Language Models
Paper • 2409.11136 • Published • 21 -
On the Diagram of Thought
Paper • 2409.10038 • Published • 12 -
Video Instruction Tuning With Synthetic Data
Paper • 2410.02713 • Published • 38 -
Large Language Models as Markov Chains
Paper • 2410.02724 • Published • 30 -
Contrastive Localized Language-Image Pre-Training
Paper • 2410.02746 • Published • 33 -
Training Language Models on Synthetic Edit Sequences Improves Code Synthesis
Paper • 2410.02749 • Published • 12 -
L-CiteEval: Do Long-Context Models Truly Leverage Context for Responding?
Paper • 2410.02115 • Published • 10 -
Interpreting and Editing Vision-Language Representations to Mitigate Hallucinations
Paper • 2410.02762 • Published • 9 -
Layer Swapping for Zero-Shot Cross-Lingual Transfer in Large Language Models
Paper • 2410.01335 • Published • 5 -
RATIONALYST: Pre-training Process-Supervision for Improving Reasoning
Paper • 2410.01044 • Published • 34 -
Not All LLM Reasoners Are Created Equal
Paper • 2410.01748 • Published • 28 -
Quantifying Generalization Complexity for Large Language Models
Paper • 2410.01769 • Published • 13 -
InfiniPot: Infinite Context Processing on Memory-Constrained LLMs
Paper • 2410.01518 • Published • 2 -
Law of the Weakest Link: Cross Capabilities of Large Language Models
Paper • 2409.19951 • Published • 53 -
Hyper-Connections
Paper • 2409.19606 • Published • 20 -
Instruction Following without Instruction Tuning
Paper • 2409.14254 • Published • 27 -
LongGenBench: Long-context Generation Benchmark
Paper • 2410.04199 • Published • 17 -
Erasing Conceptual Knowledge from Language Models
Paper • 2410.02760 • Published • 13 -
Differential Transformer
Paper • 2410.05258 • Published • 168 -
LLMs Know More Than They Show: On the Intrinsic Representation of LLM Hallucinations
Paper • 2410.02707 • Published • 47 -
Addition is All You Need for Energy-efficient Language Models
Paper • 2410.00907 • Published • 144 -
Selective Attention Improves Transformer
Paper • 2410.02703 • Published • 23 -
Mentor-KD: Making Small Language Models Better Multi-step Reasoners
Paper • 2410.09037 • Published • 4 -
Rethinking Data Selection at Scale: Random Selection is Almost All You Need
Paper • 2410.09335 • Published • 16 -
StructRAG: Boosting Knowledge Intensive Reasoning of LLMs via Inference-time Hybrid Information Structurization
Paper • 2410.08815 • Published • 43 -
SuperCorrect: Supervising and Correcting Language Models with Error-Driven Insights
Paper • 2410.09008 • Published • 16 -
Mechanistic Permutability: Match Features Across Layers
Paper • 2410.07656 • Published • 16 -
SimpleStrat: Diversifying Language Model Generation with Stratification
Paper • 2410.09038 • Published • 4 -
PositionID: LLMs can Control Lengths, Copy and Paste with Explicit Positional Awareness
Paper • 2410.07035 • Published • 16 -
ProSA: Assessing and Understanding the Prompt Sensitivity of LLMs
Paper • 2410.12405 • Published • 13 -
Exploring Model Kinship for Merging Large Language Models
Paper • 2410.12613 • Published • 19 -
Your Mixture-of-Experts LLM Is Secretly an Embedding Model For Free
Paper • 2410.10814 • Published • 48 -
What Matters in Transformers? Not All Attention is Needed
Paper • 2406.15786 • Published • 29 -
Vector-ICL: In-context Learning with Continuous Vector Representations
Paper • 2410.05629 • Published • 3 -
Intriguing Properties of Large Language and Vision Models
Paper • 2410.04751 • Published • 16 -
AutoTrain: No-code training for state-of-the-art models
Paper • 2410.15735 • Published • 58 -
Pre-training Distillation for Large Language Models: A Design Space Exploration
Paper • 2410.16215 • Published • 15 -
In-context learning and Occam's razor
Paper • 2410.14086 • Published • 2 -
SeerAttention: Learning Intrinsic Sparse Attention in Your LLMs
Paper • 2410.13276 • Published • 25 -
How Do Training Methods Influence the Utilization of Vision Models?
Paper • 2410.14470 • Published • 4 -
Context is Key(NMF): Modelling Topical Information Dynamics in Chinese Diaspora Media
Paper • 2410.12791 • Published • 4 -
Counting Ability of Large Language Models and Impact of Tokenization
Paper • 2410.19730 • Published • 10 -
Analysing the Residual Stream of Language Models Under Knowledge Conflicts
Paper • 2410.16090 • Published • 7 -
Fictitious Synthetic Data Can Improve LLM Factuality via Prerequisite Learning
Paper • 2410.19290 • Published • 10 -
On Memorization of Large Language Models in Logical Reasoning
Paper • 2410.23123 • Published • 18 -
Toxicity of the Commons: Curating Open-Source Pre-Training Data
Paper • 2410.22587 • Published • 10 -
Zero-Shot Dense Retrieval with Embeddings from Relevance Feedback
Paper • 2410.21242 • Published • 8 -
Task Vectors are Cross-Modal
Paper • 2410.22330 • Published • 11 -
RARe: Retrieval Augmented Retrieval with In-Context Examples
Paper • 2410.20088 • Published • 5 -
LongReward: Improving Long-context Large Language Models with AI Feedback
Paper • 2410.21252 • Published • 17 -
TokenFormer: Rethinking Transformer Scaling with Tokenized Model Parameters
Paper • 2410.23168 • Published • 24 -
Constraint Back-translation Improves Complex Instruction Following of Large Language Models
Paper • 2410.24175 • Published • 16 -
Language Models can Self-Lengthen to Generate Long Texts
Paper • 2410.23933 • Published • 17 -
What Happened in LLMs Layers when Trained for Fast vs. Slow Thinking: A Gradient Perspective
Paper • 2410.23743 • Published • 59 -
LoRA-Contextualizing Adaptation of Large Multimodal Models for Long Document Understanding
Paper • 2411.01106 • Published • 4 -
Physics in Next-token Prediction
Paper • 2411.00660 • Published • 14 -
GPT or BERT: why not both?
Paper • 2410.24159 • Published • 14 -
Decoding Dark Matter: Specialized Sparse Autoencoders for Interpreting Rare Concepts in Foundation Models
Paper • 2411.00743 • Published • 6 -
Needle Threading: Can LLMs Follow Threads through Near-Million-Scale Haystacks?
Paper • 2411.05000 • Published • 21 -
Analyzing The Language of Visual Tokens
Paper • 2411.05001 • Published • 22 -
Both Text and Images Leaked! A Systematic Analysis of Multimodal LLM Data Contamination
Paper • 2411.03823 • Published • 43 -
DELIFT: Data Efficient Language model Instruction Fine Tuning
Paper • 2411.04425 • Published • 9 -
The Semantic Hub Hypothesis: Language Models Share Semantic Representations Across Languages and Modalities
Paper • 2411.04986 • Published • 5 -
Counterfactual Generation from Language Models
Paper • 2411.07180 • Published • 5 -
Cut Your Losses in Large-Vocabulary Language Models
Paper • 2411.09009 • Published • 43 -
Large Language Models Can Self-Improve in Long-context Reasoning
Paper • 2411.08147 • Published • 62 -
Can sparse autoencoders be used to decompose and interpret steering vectors?
Paper • 2411.08790 • Published • 8 -
M-Longdoc: A Benchmark For Multimodal Super-Long Document Understanding And A Retrieval-Aware Tuning Framework
Paper • 2411.06176 • Published • 44 -
Top-nσ: Not All Logits Are You Need
Paper • 2411.07641 • Published • 18 -
Drowning in Documents: Consequences of Scaling Reranker Inference
Paper • 2411.11767 • Published • 17 -
Multimodal Autoregressive Pre-training of Large Vision Encoders
Paper • 2411.14402 • Published • 43 -
OpenScholar: Synthesizing Scientific Literature with Retrieval-augmented LMs
Paper • 2411.14199 • Published • 29 -
Do I Know This Entity? Knowledge Awareness and Hallucinations in Language Models
Paper • 2411.14257 • Published • 9 -
Patience Is The Key to Large Language Model Reasoning
Paper • 2411.13082 • Published • 7 -
Loss-to-Loss Prediction: Scaling Laws for All Datasets
Paper • 2411.12925 • Published • 5 -
Beyond Examples: High-level Automated Reasoning Paradigm in In-Context Learning via MCTS
Paper • 2411.18478 • Published • 32 -
Training Noise Token Pruning
Paper • 2411.18092 • Published • 1 -
Star Attention: Efficient LLM Inference over Long Sequences
Paper • 2411.17116 • Published • 47 -
Predicting Emergent Capabilities by Finetuning
Paper • 2411.16035 • Published • 6 -
Truth or Mirage? Towards End-to-End Factuality Evaluation with LLM-OASIS
Paper • 2411.19655 • Published • 20 -
Free Process Rewards without Process Labels
Paper • 2412.01981 • Published • 28 -
Critical Tokens Matter: Token-Level Contrastive Estimation Enhence LLM's Reasoning Capability
Paper • 2411.19943 • Published • 55 -
Establishing Task Scaling Laws via Compute-Efficient Model Ladders
Paper • 2412.04403 • Published • 2 -
Marco-LLM: Bridging Languages via Massive Multilingual Training for Cross-Lingual Enhancement
Paper • 2412.04003 • Published • 9 -
Densing Law of LLMs
Paper • 2412.04315 • Published • 16 -
Evaluating Language Models as Synthetic Data Generators
Paper • 2412.03679 • Published • 43 -
If You Can't Use Them, Recycle Them: Optimizing Merging at Scale Mitigates Performance Tradeoffs
Paper • 2412.04144 • Published • 4 -
KaSA: Knowledge-Aware Singular-Value Adaptation of Large Language Models
Paper • 2412.06071 • Published • 7 -
I Don't Know: Explicit Modeling of Uncertainty with an [IDK] Token
Paper • 2412.06676 • Published • 9 -
Learned Compression for Compressed Learning
Paper • 2412.09405 • Published • 11 -
Multimodal Latent Language Modeling with Next-Token Diffusion
Paper • 2412.08635 • Published • 41 -
Smaller Language Models Are Better Instruction Evolvers
Paper • 2412.11231 • Published • 25 -
RetroLLM: Empowering Large Language Models to Retrieve Fine-grained Evidence within Generation
Paper • 2412.11919 • Published • 33 -
No More Adam: Learning Rate Scaling at Initialization is All You Need
Paper • 2412.11768 • Published • 41 -
Emergence of Abstractions: Concept Encoding and Decoding Mechanism for In-Context Learning in Transformers
Paper • 2412.12276 • Published • 15 -
Are Your LLMs Capable of Stable Reasoning?
Paper • 2412.13147 • Published • 87 -
AntiLeak-Bench: Preventing Data Contamination by Automatically Constructing Benchmarks with Updated Real-World Knowledge
Paper • 2412.13670 • Published • 4 -
Smarter, Better, Faster, Longer: A Modern Bidirectional Encoder for Fast, Memory Efficient, and Long Context Finetuning and Inference
Paper • 2412.13663 • Published • 103