πŸ““ Note Repository
  • β–Ά πŸ“ Competitions
    • β–Ά πŸ“ ARC-WhiteBox-Estimation Challenge
      • πŸ“„ Challenge Details
    • β–Ά πŸ“ TrexQuant Alpha Challenge
      • πŸ“„ Alpha Info
    • πŸ“„ LPCV-Track-1
    • πŸ“„ Orbit Wars
  • β–Ά πŸ“ Courses
    • β–Ά πŸ“ FM and Diff - MIT CSAIL
      • πŸ“„ Lecture-1
      • πŸ“„ Lecture-2
      • πŸ“„ Overview
    • β–Ά πŸ“ Signal Processing
      • β–Ά πŸ“ Digital Communication
        • πŸ“„ Untitled
      • β–Ά πŸ“ Fundamentals
        • πŸ“„ Fourier Transform
  • β–Ά πŸ“ Deep Learning
    • πŸ“„ Attention Mechanism
    • πŸ“„ Continual Learning
    • πŸ“„ Continual Learning - Eval Metrics
    • πŸ“„ Continual Learning Formulation
    • πŸ“„ Curriculum Learning
    • πŸ“„ Data Attribution
    • πŸ“„ Data-Driven Inductive Bias
    • πŸ“„ Depth Estimation Metrics
    • πŸ“„ Deterministic vs Probabilistic Models
    • πŸ“„ Different ML Optimizers
    • πŸ“„ Diffusion Models
    • πŸ“„ Diffusion Reverse Process
    • πŸ“„ Diffusion-LM Denoising
    • πŸ“„ Energy based Models
    • πŸ“„ FrΓ©chet Inception Distance (FID)
    • πŸ“„ Low Rank Adaptation(LoRA)
    • πŸ“„ Matryoshka Representation Learning
    • πŸ“„ Mixture of Experts
    • πŸ“„ ML Notes Database Expansion
    • πŸ“„ Moravec's Paradox
    • πŸ“„ N-gram Models
    • πŸ“„ Neural Implicit Fields
    • πŸ“„ Optimal Transport
    • πŸ“„ Predicting Next Word with GPT-2
    • πŸ“„ Reimannian Manifolds
    • πŸ“„ Solvable Models in Machine Learning
    • πŸ“„ Stability vs Plasticity
    • πŸ“„ The Manifold Hypothesis
    • πŸ“„ VQ-VAE (Vector Quantized Variational Autoencoder)
    • πŸ“„ What is MoCo?
    • πŸ“„ What is Perplexity in ML?
    • πŸ“„ x_0-parameterization
  • β–Ά πŸ“ DSA and CP
    • β–Ά πŸ“ Binary Search
      • πŸ“„ Find 1st and last pos of element in sorted array
      • πŸ“„ Find Peak Element
      • πŸ“„ Koko eating Banana
      • πŸ“„ Note on Binary Search Methods
      • πŸ“„ Search in 2d - II
      • πŸ“„ Search in a 2d array
      • πŸ“„ Search in Rotated Array
      • πŸ“„ Single Element in a Sorted array of pairs
      • πŸ“„ Split Array Largest Sum
    • β–Ά πŸ“ Binary Trees
      • πŸ“„ Binary Search Trees
      • πŸ“„ Binary Trees
      • πŸ“„ Check for a Balanced BT
      • πŸ“„ Inorder Traversal of a BT
      • πŸ“„ Postorder Traversal of BT
      • πŸ“„ Preorder Inorder Postorder Traversals in One Traversal
    • β–Ά πŸ“ Doubts
      • πŸ“„ Binary Search β€” `<=` vs `<` Boundary Conditions
    • β–Ά πŸ“ Dynamic Programming
      • πŸ“„ Coin Change
      • πŸ“„ Subset sum equal to target
      • πŸ“„ Target Sum
      • πŸ“„ Unbounded Knapsack
    • β–Ά πŸ“ Graphs
      • πŸ“„ Connected Components in a Graph
      • πŸ“„ Course Schedule
      • πŸ“„ Course Schedule II
      • πŸ“„ Detection of a Cycle in a Directed Graph
      • πŸ“„ Detection of Cycle in an Undirected Graph
      • πŸ“„ Nearest 0 from cell in a matrix
      • πŸ“„ Rotten Oranges
      • πŸ“„ Topological Sort
    • β–Ά πŸ“ Greedy
      • πŸ“„ Insert Interval
      • πŸ“„ Jump-II
      • πŸ“„ Task Scheduler
    • β–Ά πŸ“ Heaps
      • πŸ“„ Heaps (Priority Queues) β€” Detailed Notes
    • β–Ά πŸ“ Maths
      • πŸ“„ Sieve of Eratosthenes
      • πŸ“„ Single Number Others Twice
    • β–Ά πŸ“ Sliding WIndow and Two Pointers
      • πŸ“„ Defuse the Bomb
      • πŸ“„ Fruits Into Basket
      • πŸ“„ Max Consecutive One III
      • πŸ“„ Maximum Erasure Value
    • β–Ά πŸ“ Sorting
      • πŸ“„ Merge Sorted Array
      • πŸ“„ Quick Sort
      • πŸ“„ Quick Sort Partition
    • β–Ά πŸ“ Strings
      • πŸ“„ CPP Strings CheatSheet
      • πŸ“„ Largest Odd Number in String
    • πŸ“„ Linked List
  • β–Ά πŸ“ Machine Learning
    • πŸ“„ Bias and Variance Trade-off
    • πŸ“„ Boosting
    • πŸ“„ Byte Pair Encoding (BPE) Tokenization
    • πŸ“„ Decision Trees (A comprehensive Note)
    • πŸ“„ K-Means Clustering
    • πŸ“„ Long Short-Term Memory (LSTM)
    • πŸ“„ Principal Component Analysis (PCA)
    • πŸ“„ Random Forests
    • πŸ“„ Recurrent Neural Netoworks(RNNs)
    • πŸ“„ Representational Similarity Analysis (RSA)
    • πŸ“„ t-SNE (t-Distributed Stochastic Neighbor Embedding)
    • πŸ“„ Training Stabilization in Machine Learning
    • πŸ“„ Word2Vec
  • β–Ά πŸ“ Math
    • β–Ά πŸ“ Linear Algebra
      • πŸ“„ Connection between SVD and EVD
      • πŸ“„ Singular Value Decomposition(SVD)
      • πŸ“„ Spectral Norm
    • β–Ά πŸ“ Probability Theory
      • πŸ“„ Convergences in Probability vs Almost Sure
      • πŸ“„ Deterministic Random Variable
      • πŸ“„ Gaussian Distributions
      • πŸ“„ Independent Random Variables
      • πŸ“„ Jointly Gaussian Random Variables β€” Conditional Distributions
      • πŸ“„ Law of Iterated Expectations
      • πŸ“„ Markov, Chebyshev, and Jensen's Inequalities
      • πŸ“„ p-values
      • πŸ“„ Power of Test Statistic
      • πŸ“„ Probability and Statistics
      • πŸ“„ Study Materials
      • πŸ“„ Total Expectation Theorem
      • πŸ“„ Unique Properties of Gaussian RVs
      • πŸ“„ Well Known Distributions
      • πŸ“„ Ξ± (Significance Level)
    • πŸ“„ What is Kernel(function)?
  • β–Ό πŸ“ Papers
    • β–Ά πŸ“ World Models Reading List
      • πŸ“„ I-JEPA
      • πŸ“„ What to Read
    • πŸ“„
    • πŸ“„ A Compression Perspective on Simplicity Bias
    • πŸ“„ Beyond Neural Scaling Law - Data Pruning improve power law
    • πŸ“„ Boosting Data-Driven Mirror Descent with Randomization,Equivariance, and Acceleration
    • πŸ“„ Cram Less to Fit More
    • πŸ“„ Crowd-Aware Robot Navigation with Switching BetweenLearning-Based and Rule-Based Methods Using Normalizing Flows
    • πŸ“„ Deep Learning on a Data Diet
    • πŸ“„ Diffusion-LM
    • πŸ“„ Information Theory for Repr Learning
    • πŸ“„ Invariant Risk Minimization
    • πŸ“„ Large-Scale Pruning using Dyna Uncertainty
    • πŸ“„ Learned feature representations are biased by complexity, learning order, position, and more
    • πŸ“„ Learning Stable Normalizing-Flow Control for Robotic Manipulation
    • πŸ“„ Learning to Theorise
    • πŸ“„ Learning2Optimize
    • πŸ“„ MACO-Feature Visualization via Magnitude Constrained Optimization
    • πŸ“„ MaxEnRL via Energy-based Normalising Flows 1
    • πŸ“„ Mechanistic Mode Connectivity
    • πŸ“„ Neural Ordinary Differential Equations
    • πŸ“„ RecoveryChaining-Learning Local Recovery Policies for Robust Manipulation
    • πŸ“„ RESPO (Reachability Estimation for Safe Policy Optimization)
    • πŸ“„ Scaling for Data Filtering
    • πŸ“„ TIC-CLIP - continual training of clip models
    • πŸ“„ Understanding BlackBox Predictions - Using Influence Functions
    • πŸ“„ Understanding Visual Feature Reliance through the Lens of Complexity
    • πŸ“„ When Do Curricula Work?
  • β–Ά πŸ“ Projects
    • β–Ά πŸ“ AlphaGo-Black-Hole
      • πŸ“„ Alpha-Beta Algorithm (Chess Engines)
      • πŸ“„ AlphaZero and Chess
      • πŸ“„ AlphaZero Training
      • πŸ“„ Alphazero Workflow
      • πŸ“„ Improved Representations
      • πŸ“„ Mixture of Experts Approach to Games
    • β–Ά πŸ“ DLM-Revisited
      • πŸ“„ Workbench
  • πŸ“„ Compositional Understanding leads to Causal
  • πŸ“„ Cool Things I Read
  • πŸ“„ DIFFUSION POSTERIOR SAMPLING FOR GENERAL NOISY INVERSE PROBLEMS
  • πŸ“„ Explaining Data-Mixing Scaling Laws
  • πŸ“„ Fun Resources
  • πŸ“„ Hand of Straights
  • πŸ“„ House Robber II
  • πŸ“„ How Feature Learning can Improve Neural Scaling
  • πŸ“„ Longest common Prefix and Sorting a String
  • πŸ“„ Not all solutions are created equal
  • πŸ“„ Professors and Labs
  • πŸ“„ Untitled
  • πŸ“„ Untitled 1
  • πŸ“„ Untitled 2
  • πŸ“„ Untitled 3

Utsab's Note Repository

Cram Less to Fit More

Last modified: July 21, 2026

#nlp #model_capacity #scaling-laws source: \[2604.08519\] Cram Less to Fit More: Training Data Pruning Improves Memorization of Facts

© 2026 Utsab's Note Repository. Built with Hugo.