BestHub
Sign in
Collection size
94 articles
Page 2 of 5
Xiaohongshu Tech REDtech
Xiaohongshu Tech REDtech
Aug 5, 2026 · Artificial Intelligence

Accelerating Xiaohongshu ‘Ask’ Inference: Slim Vision Tokens, Focused MoE

The article details how Xiaohongshu’s multimodal “Ask” service reduces visual token bloat and MoE expert compute by applying dynamic Vision Token compression (VisionZip) and similarity‑based expert re‑routing (SERE), achieving up to 13% lower first‑token latency and 16.5% faster end‑to‑end throughput while preserving answer quality.

MoE expert routingQwen modelsSERE
0 likes · 19 min read
Accelerating Xiaohongshu ‘Ask’ Inference: Slim Vision Tokens, Focused MoE
Baobao Algorithm Notes
Baobao Algorithm Notes
Apr 27, 2025 · Artificial Intelligence

How DeepSeek R1T‑Chimera Cuts Tokens by 40% Without Fine‑Tuning

The DeepSeek‑R1T‑Chimera model merges DeepSeek‑R1 reasoning with V3‑0324 architecture, reusing most V3 weights and swapping only the blue‑highlighted R1 routing experts, achieving the same intelligence as R1 while reducing output tokens by about 40% and running faster, all without any fine‑tuning or distillation.

DeepSeekLLMParameter Reuse
0 likes · 5 min read
How DeepSeek R1T‑Chimera Cuts Tokens by 40% Without Fine‑Tuning
Xiaohongshu Tech REDtech
Xiaohongshu Tech REDtech
Nov 4, 2025 · Artificial Intelligence

Unveiling the Law of Capacity Gap: Boosting Language Model Distillation Efficiency

At ACL 2025, a collaborative paper introduced the Law of Capacity Gap, revealing a linear 2.5× optimal teacher‑student size relationship in language model distillation, dramatically cutting compute costs and achieving Pareto‑optimal efficiency, with the MiniMA model as a successful demonstration.

MiniMAartificial-intelligencecapacity-gap
0 likes · 7 min read
Unveiling the Law of Capacity Gap: Boosting Language Model Distillation Efficiency
AntTech
AntTech
Jun 21, 2025 · Artificial Intelligence

Ring-lite: Open‑Source Lightweight MoE Model Sets SOTA on AIME and LiveCodeBench

Ring-lite, an open‑source lightweight Mixture‑of‑Experts inference model built on Ling‑lite‑1.5, introduces the C3PO reinforcement‑learning training method and achieves state‑of‑the‑art results on benchmarks such as AIME24/25, LiveCodeBench, CodeForce, and GPQA‑diamond, while offering full transparency of weights, code, and data.

AI inferenceC3PObenchmark
0 likes · 11 min read
Ring-lite: Open‑Source Lightweight MoE Model Sets SOTA on AIME and LiveCodeBench
Didi Tech
Didi Tech
Mar 12, 2026 · Artificial Intelligence

How STAPO Improves Large‑Model Fine‑Tuning by Silencing Spurious Tokens

The STAPO (Spurious‑Token‑Aware Policy Optimization) algorithm, introduced by Tsinghua University's iDLab and Didi's Deep Sea Lab, tackles policy‑entropy instability and performance oscillation in reinforcement‑learning fine‑tuning of large models by mathematically analyzing token collision probability, defining spurious tokens, and applying a Silencing Spurious Tokens mechanism that yields state‑of‑the‑art results on multiple math‑reasoning benchmarks.

AI safetySTAPOfine-tuning
0 likes · 7 min read
How STAPO Improves Large‑Model Fine‑Tuning by Silencing Spurious Tokens
Old Zhang's AI Learning
Old Zhang's AI Learning
Apr 26, 2026 · Artificial Intelligence

Distilling Claude Opus into Qwen3.6-27B – GGUF Lets You Run Locally on Consumer GPUs

The preview model Qwopus3.6-27B‑v1, distilled from Claude Opus onto Qwen3.6‑27B using SFT with the Unsloth stack and a curated 12 K high‑quality inference sample set, is evaluated on agentic reasoning, front‑end design, and Canvas/WebGL tasks with an RTX 5090, and can be deployed locally via llama.cpp GGUF quantizations with detailed memory guidelines.

Apache 2.0Claude OpusGGUF
0 likes · 7 min read
Distilling Claude Opus into Qwen3.6-27B – GGUF Lets You Run Locally on Consumer GPUs
Alibaba Cloud Big Data AI Platform
Alibaba Cloud Big Data AI Platform
Mar 29, 2025 · Artificial Intelligence

How DistilQwen2.5‑R1 Boosts Small‑Model Reasoning with Innovative Knowledge Distillation

The article introduces the DistilQwen2.5‑R1 series, which leverages a novel knowledge‑distillation pipeline—including CoT data evaluation, improvement, and validation—to transfer deep reasoning abilities from large models like DeepSeek‑R1 to compact models, achieving superior performance across math, code, and scientific benchmarks and providing open‑source checkpoints and deployment guides for practical use.

AI inferenceLarge Language Modelsbenchmark evaluation
0 likes · 17 min read
How DistilQwen2.5‑R1 Boosts Small‑Model Reasoning with Innovative Knowledge Distillation
Architect Practice
Architect Practice
Jul 29, 2026 · Artificial Intelligence

How a 1.5B Model Beats Cutting‑Edge Large Models on Math Exams

This article explains why and how knowledge distillation lets a 1.5 B parameter model surpass much larger LLMs on math benchmarks, detailing the underlying soft‑label transfer, temperature tuning, various distillation families, engineering pipelines, and the practical trade‑offs that bound its success.

AI EngineeringLarge Language Modelsknowledge distillation
0 likes · 14 min read
How a 1.5B Model Beats Cutting‑Edge Large Models on Math Exams
AntTech
AntTech
Apr 23, 2026 · Artificial Intelligence

Ling-2.6-flash: Faster Response, Stronger Execution, and Higher Token Efficiency for Agent Workloads

Ling-2.6-flash is a 104B‑parameter Instruct model that uses a mixed‑linear architecture and token‑efficiency optimizations to achieve up to 340 tokens/s inference speed, 4× higher throughput than comparable models, and ten‑fold lower token consumption on Agent benchmarks, while maintaining SOTA performance.

Agent OptimizationLLMbenchmark
0 likes · 15 min read
Ling-2.6-flash: Faster Response, Stronger Execution, and Higher Token Efficiency for Agent Workloads
DataFunTalk
DataFunTalk
Oct 30, 2025 · Artificial Intelligence

How On-Policy Distillation Cuts LLM Training Cost by 90%

Thinking Machines Lab introduces On-Policy Distillation, a post‑training technique that matches reinforcement‑learning performance while reducing compute cost by up to tenfold, and demonstrates its effectiveness through extensive experiments on reasoning, personalization, and catastrophic‑forgetting mitigation.

Model EfficiencyOn-Policy Distillationknowledge distillation
0 likes · 15 min read
How On-Policy Distillation Cuts LLM Training Cost by 90%
360 Tech Engineering
360 Tech Engineering
Aug 7, 2026 · Artificial Intelligence

Token Compression: From Simple Text Trimming to LLM Context Governance

The article explains how token compression evolves from basic text shortening into a multi‑layered context‑governance process for large language models, balancing compression rate, semantic fidelity, constraint integrity, efficiency, stability and observability while deciding when and how to apply it.

LLM contextcontext managementcost optimization
0 likes · 20 min read
Token Compression: From Simple Text Trimming to LLM Context Governance
ThinkingAgent
ThinkingAgent
Jun 17, 2026 · Artificial Intelligence

Why a 7B Model Can Outperform a 70B Model: The Power of Knowledge Distillation

The article explains how knowledge distillation lets a small LLM learn from a much larger teacher model, achieving near‑teacher performance while cutting inference cost, latency and memory, and provides a step‑by‑step guide, benchmark results, advanced on‑policy techniques, common pitfalls and best‑practice recommendations.

LLMLLaMA-3On-Policy Distillation
0 likes · 13 min read
Why a 7B Model Can Outperform a 70B Model: The Power of Knowledge Distillation
Top Architect
Top Architect
Feb 14, 2025 · Artificial Intelligence

DeepSeek Model Distillation: Principles, Innovations, Architecture, and Performance

This article provides an in‑depth overview of DeepSeek’s model distillation technology, covering its definition, core principles, innovative data‑model distillation integration, architecture design, training strategies, performance gains, and the challenges of scaling to multimodal data.

DeepSeekLarge Language Modelsai-optimization
0 likes · 16 min read
DeepSeek Model Distillation: Principles, Innovations, Architecture, and Performance
Baobao Algorithm Notes
Baobao Algorithm Notes
Jun 28, 2024 · Artificial Intelligence

What Makes Gemma 2 a Competitive Open‑Source LLM? Architecture, Training, and Evaluation Insights

The article provides a detailed technical overview of Gemma 2, covering its decoder‑only transformer design, novel attention mechanisms, logit soft‑capping, RMSNorm, knowledge‑distillation training on trillions of tokens, extensive pre‑training infrastructure, and benchmark evaluations that demonstrate its competitiveness against larger proprietary models.

AIGemma 2benchmark evaluation
0 likes · 14 min read
What Makes Gemma 2 a Competitive Open‑Source LLM? Architecture, Training, and Evaluation Insights
Machine Learning Algorithms & Natural Language Processing
Machine Learning Algorithms & Natural Language Processing
Oct 19, 2026 · Artificial Intelligence

AdaFlash: Adaptive Speculative Decoding with Online Self-Improvement for High-Concurrency LLM Inference

Nanjing University and Huawei's Foundation Model team introduce AdaFlash, an adaptive speculative decoding framework that enables diffusion draft models to self-improve online via on-policy distillation and dynamic verification length prediction, achieving up to 5.32x end-to-end speedup on Qwen3-8B and 66% higher throughput under high concurrency.

AdaFlashLLM inference accelerationOn-Policy Distillation
0 likes · 17 min read
AdaFlash: Adaptive Speculative Decoding with Online Self-Improvement for High-Concurrency LLM Inference
AntTech
AntTech
Mar 11, 2024 · Artificial Intelligence

Can Small Language Models be Good Reasoners in Recommender Systems?

This article presents SLIM, a knowledge‑distillation framework that transfers the reasoning abilities of large language models to compact models for sequential recommendation, enhancing item representation, user profiling, and bias mitigation while achieving comparable performance with far lower computational resources.

AIEfficiencyLLM
0 likes · 12 min read
Can Small Language Models be Good Reasoners in Recommender Systems?
Architect
Architect
Feb 9, 2025 · Artificial Intelligence

How DeepSeek’s Model Distillation Boosts AI Efficiency and Performance

This article provides an in‑depth analysis of DeepSeek’s model distillation technology, covering its definition, core principles, innovative strategies, architecture design, training optimizations, benchmark results, efficiency gains, and the remaining challenges of applying distillation to large language models and multimodal data.

AI EfficiencyDeepSeekLarge Language Models
0 likes · 16 min read
How DeepSeek’s Model Distillation Boosts AI Efficiency and Performance