Tagged articles
KL Divergence
18 articles · Page 1 of 1
Soft-MTP decomposes the cross-entropy gradient in multi-token prediction into a distribution-alignment term and a residual term, then cancels the residual from gradients flowing to MTP heads and the shared backbone while preserving cross-entropy training on the shared output projection, improving language modeling and speculative decoding without extra parameters or full KL computation.
Cross-Entropy LossGradient DecompositionKL Divergence
0 likes · 17 min read

This article breaks down six high-frequency GRPO interview questions from top Chinese tech companies, covering GRPO vs PPO trade-offs, group-relative advantage calculation with concrete numbers, handling all-correct/all-wrong sample groups, KL constraint mechanics, convergence monitoring priorities, and reward hacking detection via shadow evaluation.
Advantage EstimationGRPOKL Divergence
0 likes · 19 min read


Geek LabsSep 9, 2026 · Artificial Intelligence
This article explains how Heretic automates directional ablation to permanently remove refusal behavior from language models by editing model weights, contrasting it with jailbreaks and steering vectors, and details its use of Optuna optimization with KL divergence to preserve model capabilities, situating it within a broader taxonomy of model modification techniques.
AbliterationHereticKL Divergence
0 likes · 18 min read

The article explains that forward KL encourages a student model to cover all major teacher modes, whereas reverse KL seeks a single dominant mode, and shows why online reinforcement learning methods like OPD and RLHF adopt reverse KL while offline SFT distillation relies on forward KL.
KL DivergenceOPDRLHF
0 likes · 7 min read

The LoopCoder‑v2 study shows that Parallel Loop Transformers achieve their best code‑generation performance with two refinement loops, as additional loops increase memory cost without improving results and even cause performance degradation, a finding explained through detailed metric analysis and cost‑benefit reasoning.
G-SWAKL DivergenceLoopCoder-v2
0 likes · 14 min read

The paper shows that SFT, DAgger, offline RL and OPD are the four orthogonal combinations of prefix source (teacher vs. student) and KL direction (forward vs. reverse), exposing three hidden trade‑offs—KL direction, prefix source, and training length—and proposes KL‑mixing and entropy‑gated length curricula that boost Avg@k by 3.6 points, raise Pass@k by up to 5.8 points, and cut response length by three‑fold.
DAggerKL DivergenceLLM distillation
0 likes · 17 min read

DiffusionOPD introduces a unified on‑policy distillation framework for diffusion models that decouples single‑task online policy exploration from multi‑task capability integration, training expert teachers per task and distilling their skills into a single student model, achieving faster convergence and higher performance across composition, OCR, and aesthetic tasks.
KL DivergencePPOReinforcement Learning
0 likes · 8 min read

The article explains the definitions, properties, and asymmetric nature of KL divergence, compares Forward KL (mean‑seeking) and Reverse KL (mode‑seeking) through bimodal examples, and provides practical guidelines for choosing between them based on sampling and probability‑evaluation capabilities in machine‑learning tasks.
KL Divergenceforward KLmachine learning
0 likes · 10 min read

The article explains On-Policy Distillation, derives its forward and reverse KL gradients, introduces Self‑Distillation where the policy serves as its own teacher, discusses practical implementation tricks such as extra‑knowledge injection, EMA or trust‑region teacher stabilization, and highlights benefits like reduced catastrophic forgetting, fewer Aha moments, and a narrower train‑test gap, especially for larger models.
Catastrophic ForgettingEMAKL Divergence
0 likes · 6 min read

The LightReasoner paper from Hong Kong University shows that small language models can guide large models on critical reasoning steps, achieving up to 90% faster inference and significant accuracy gains across multiple math benchmarks.
Contrastive DecodingKL DivergenceMathematical Reasoning
0 likes · 9 min read

This article explains the Kullback‑Leibler (KL) divergence for discrete and continuous distributions, outlines its non‑negativity and asymmetry, walks through a uniform‑distribution example, provides a simple Python demonstration, and discusses key applications in variational autoencoders, reinforcement‑learning policy optimization, and other machine‑learning contexts.
KL DivergenceVariational AutoEncoderinformation theory
0 likes · 7 min read

This article explains knowledge distillation as used in DeepSeek's R1 model, detailing how a large teacher model transfers knowledge to a smaller student model via soft targets and loss functions like KL divergence, enabling efficient deployment on resource-constrained devices.
AI EfficiencyCross-Entropy LossDeepSeek-R1
0 likes · 6 min read

The article explains Shannon information content, entropy, cross‑entropy, KL‑divergence, conditional entropy and mutual information, illustrating each concept with coin‑flip and dice examples, visual formulas, and discusses their roles as loss functions and evaluation metrics in machine‑learning models.
KL Divergencecross-entropyentropy
0 likes · 8 min read


Code DAODec 20, 2021 · Artificial Intelligence
This article explains the theory behind variational autoencoders, details their KL‑divergence loss, provides a complete TensorFlow implementation, and demonstrates reconstruction, latent‑space visualization, and novel image generation through sampling and interpolation.
Image GenerationKL DivergencePython
0 likes · 13 min read


Code DAODec 10, 2021 · Artificial Intelligence
This article explains the principles of variational autoencoders, starting with dimensionality reduction techniques such as PCA and standard autoencoders, highlighting their limitations for data generation, and then detailing VAE's regularized latent space, variational inference, re‑parameterization, and loss formulation.
Generative ModelsKL DivergenceVAE
0 likes · 18 min read


21CTOFeb 7, 2018 · Artificial Intelligence
This article explains entropy, joint entropy, conditional entropy, and related measures such as KL divergence and cross‑entropy, using intuitive coin‑flip examples and mathematical formulas to show how they quantify uncertainty and information in probability distributions.
KL Divergencecross-entropyentropy
0 likes · 14 min read

This article explains the concepts of entropy, joint entropy, conditional entropy, relative entropy (KL divergence) and cross‑entropy, illustrating their definitions, mathematical formulas, intuitive interpretations, and relationships through simple probability examples and visual diagrams.
KL Divergencecross-entropyentropy
0 likes · 14 min read

This article reviews the most frequently used distance and similarity formulas in machine learning and data mining, explaining their definitions, mathematical properties, practical examples, and when each metric is appropriate for measuring differences between data points or probability distributions.
KL DivergenceMahalanobis distancecosine similarity
0 likes · 13 min read
