SOTAVerified

Memorization

Papers

Recently AddedMost HypedMost ActiveNeeds VerificationMost Verified

Showing 1–50 of 1088 papers

TitleStatusHype
What Should LLMs Forget? Quantifying Personal Data in LLMs for Right-to-Be-Forgotten Requests—0
Reasoning or Memorization? Unreliable Results of Reinforcement Learning Due to Data ContaminationCode1
Entropy-Memorization Law: Evaluating Memorization Difficulty of Data in LLMs—0
MMReason: An Open-Ended Multi-Modal Multi-Step Reasoning Benchmark for MLLMs Toward AGICode0
Listener-Rewarded Thinking in VLMs for Image Preferences—0
Where to find Grokking in LLM Pretraining? Monitor Memorization-to-Generalization without Test—0
Leaner Training, Lower Leakage: Revisiting Memorization in LLM Fine-Tuning with LoRA—0
Counterfactual Influence as a Distributional Quantity—0
Uncovering Conceptual Blindspots in Generative Image Models Using Sparse Autoencoders—0
A Random Matrix Analysis of In-context Memorization for Nonlinear Attention—0
Robots and Children that Learn Together : Improving Knowledge Retention by Teaching Peer-Like Interactive Robots—0
In-Context Learning Strategies Emerge Rationally—0
Winter Soldier: Backdooring Language Models at Pre-Training with Indirect Data Poisoning—0
Capacity Matters: a Proof-of-Concept for Transformer Memorization on Real-World DataCode0
Dataset distillation for memorized data: Soft labels can leak held-out teacher knowledgeCode0
Less is More: Undertraining Experts Improves Model Upcycling—0
LexiMark: Robust Watermarking via Lexical Substitutions to Enhance Membership Verification of an LLM's Textual Training DataCode0
Sharpness-Aware Machine Unlearning—0
The SWE-Bench Illusion: When State-of-the-Art LLMs Remember Instead of Reason—0
Restoring Gaussian Blurred Face Images for Deanonymization Attacks—0
SoK: Data Reconstruction Attacks Against Machine Learning Models: Definition, Metrics, and Benchmark—0
Private Memorization Editing: Turning Memorization into a Defense to Strengthen Data Privacy in Large Language ModelsCode0
HeuriGym: An Agentic Benchmark for LLM-Crafted Heuristics in Combinatorial OptimizationCode2
Generative Modeling of Weights: Generalization or Memorization?Code1
Diffusion models under low-noise regimeCode0
Simple Yet Effective: Extracting Private Data Across Clients in Federated Fine-Tuning of Large Language Models—0
Quantifying Cross-Modality Memorization in Vision-Language Models—0
Membership Inference Attacks on Sequence Models—0
Beyond Memorization: A Rigorous Evaluation Framework for Medical Knowledge EditingCode0
Trade-offs in Data Memorization via Strong Data Processing Inequalities—0
How much do language models memorize?—0
How does Transformer Learn Implicit Reasoning?Code1
MathArena: Evaluating LLMs on Uncontaminated Math CompetitionsCode3
Bayesian Perspective on Memorization and Reconstruction—0
Benchmarking Abstract and Reasoning Abilities Through A Theoretical PerspectiveCode0
Navigating the Latent Space Dynamics of Neural Models—0
OWL: Probing Cross-Lingual Recall of Memorized Texts via World LiteratureCode0
Kernel-Smoothed Scores for Denoising Diffusion: A Bias-Variance Study—0
What is Adversarial Training for Diffusion Models?—0
Memorization to Generalization: Emergence of Diffusion Models from Associative Memory—0
Emergent LLM behaviors are observationally equivalent to data leakageCode0
Spurious Privacy Leakage in Neural Networks—0
Grokking ExPLAIND: Unifying Model, Data, and Training Attribution to Study Model BehaviorCode0
Understanding Generalization in Diffusion Models via Probability Flow Distance—0
Querying Kernel Methods Suffices for Reconstructing their Training DataCode0
Discovering Forbidden Topics in Language Models—0
Why Diffusion Models Don't Memorize: The Role of Implicit Dynamical Regularization in Training—0
Memorization or Reasoning? Exploring the Idiom Understanding of LLMs—0
Sudoku-Bench: Evaluating creative reasoning with Sudoku variantsCode0
Understanding Fact Recall in Language Models: Why Two-Stage Training Encourages Memorization but Mixed Training Teaches Knowledge—0
Show:102550
← PrevPage 1 of 22Next →

Benchmark Results

#ModelMetricClaimedVerifiedStatus
1PaLM-540B (few-shot, k=5)Accuracy95.4—Unverified
2Gopher-280B (few-shot, k=5)Accuracy80—Unverified
3PaLM-62B (few-shot, k=5)Accuracy77.7—Unverified