BestHub
Sign in
Wu Shixiong's Large Model Academy
Author

Wu Shixiong's Large Model Academy

We continuously share large‑model know‑how, helping you master core skills—LLM, RAG, fine‑tuning, deployment—from zero to job offer, tailored for career‑switchers, autumn recruiters, and those seeking stable large‑model positions.

144
Articles
0
Likes
1.2k
Views
0
Comments
Recent Articles

Latest from Wu Shixiong's Large Model Academy

100 recent articles max
Wu Shixiong's Large Model Academy
Wu Shixiong's Large Model Academy
Aug 24, 2026 · Artificial Intelligence

Why Embedding Choice Outweighs Reranker in RAG Model Selection

This article explains why embedding model selection must precede reranker evaluation in RAG systems, detailing a three-stage evaluation methodology using business-specific data to measure recall, ranking quality, and end-to-end answer validity, while accounting for engineering constraints like latency, resource usage, and failure modes.

EmbeddingLLM applicationsMTEB
0 likes · 14 min read
Why Embedding Choice Outweighs Reranker in RAG Model Selection
Wu Shixiong's Large Model Academy
Wu Shixiong's Large Model Academy
Aug 13, 2026 · Backend Development

Why asyncio.gather Errors Don’t Halt All Branches and How to Design Robust Failure Strategies

The article explains that asyncio.gather only propagates the first exception without cancelling other awaitables, distinguishes fan‑out/fan‑in from a Supervisor role, and demonstrates a pure‑Python pattern for classifying required versus optional branches, handling failures, and merging results safely.

Error HandlingFan-outPython
0 likes · 12 min read
Why asyncio.gather Errors Don’t Halt All Branches and How to Design Robust Failure Strategies
Wu Shixiong's Large Model Academy
Wu Shixiong's Large Model Academy
Aug 7, 2026 · Artificial Intelligence

Why a Single -100 Line Determines Who the Multi‑Round SFT Learns to Speak

The article explains how using -100 as an ignored label in PyTorch cross‑entropy loss silently masks non‑assistant tokens, how to locate assistant spans via prefix‑difference, the trade‑offs between supervising only the final reply versus all assistant turns, and the essential pre‑training checks to avoid hidden masking errors in multi‑round SFT.

LLMPyTorchSFT
0 likes · 14 min read
Why a Single -100 Line Determines Who the Multi‑Round SFT Learns to Speak
Wu Shixiong's Large Model Academy
Wu Shixiong's Large Model Academy
Aug 5, 2026 · Artificial Intelligence

Detecting False Promises in Customer Service Agents: Can a Large Model Score Their Claims?

The article analyzes a deterministic rule called false_promise that flags agent replies claiming completed actions without corresponding tool calls, explains how tense affects verification, proposes a four‑step “claim‑check” process, and shows how scoring caps and regression samples expose both true violations and false‑positive edge cases.

Agent VerificationCustomer Servicefalse_promise
0 likes · 10 min read
Detecting False Promises in Customer Service Agents: Can a Large Model Score Their Claims?
Wu Shixiong's Large Model Academy
Wu Shixiong's Large Model Academy
Jul 29, 2026 · Information Security

Why Claude Code Still Needs a Sandbox Even with Auto Mode

Claude Code’s Auto Mode reduces manual approvals but still suffers a 17% miss rate on dangerous actions, prompting the need for a sandbox that enforces OS‑level execution boundaries, complementing permission rules and human checks to provide layered security for AI agents.

AgentAuto ModeClaude Code
0 likes · 11 min read
Why Claude Code Still Needs a Sandbox Even with Auto Mode
Wu Shixiong's Large Model Academy
Wu Shixiong's Large Model Academy
Jul 28, 2026 · Artificial Intelligence

Why Splitting a Task into 5 Subagents Can Be Slower Than a Single Agent

The article explains that parallelizing a task across five Claude Code subagents often yields no speedup—or even slowdown—because true parallel gains depend on independent task parts, while four overheads (re‑learning context, dependency chains, shared file conflicts, and single‑point aggregation) can dominate the runtime.

Amdahl's LawClaude CodeParallelism
0 likes · 16 min read
Why Splitting a Task into 5 Subagents Can Be Slower Than a Single Agent