Reading
An open-source toolkit that uses the Claude large language model for semi-automated quantitative analysis of dream content, significantly reducing coding workload while retaining manual review.
Section 01
Introducing llm_dream_coder—an open-source toolkit that uses the Claude large language model for semi-automated Hall/Van de Castle (H/VdC) coding. It aims to solve the time-consuming and labor-intensive problem of manual coding in dream research, reducing coding workload while retaining manual review to support large-scale quantitative dream analysis.
Section 02
The H/VdC system is a standard framework for quantitative analysis of dream content, covering multiple dimensions such as characters, social interactions, and activities. Traditional manual coding requires professional training, is highly reliable but labor-intensive, limiting its application in large-scale research.
Section 03
llm_dream_coder uses a modular architecture with core principles: universality (no custom dataset required), modularity (independent coding categories), and manual review orientation (output for rechecking). Technical workflow: Data reading → API call to Claude (with coding manual prompts) → Result parsing → Evaluation and comparison → Save results, using prompt caching to reduce costs.
Section 04
Character coding module test results:
| Dataset | Type | Sample Size | Overall F1 | Non-Family Character F1 |
|---|---|---|---|---|
| b-baseline | Serial data (development set) | 50 | 0.73 | 0.74 |
| norms-f | Normative data (test set) | 50 | 0.68 | 0.70 |
| emma | Serial data (test set) | 50 | 0.51 | 0.54 |
| Key findings: Non-Family Character F1 is the core metric (family coding requires biographical knowledge); the norms dataset is the most appropriate benchmark. |
Section 05
Data preparation: Requires coded_dreams.csv (including dream_id, etc.) and optional dreambank_codings.csv. Operation modes: Default, specified quantity/set, serial mode, etc. Cost: Claude-opus-4-6 model costs approximately $0.02-$0.05 per dream; caching mechanism can reduce costs.
Section 06
Limitations: Family coding requires biographical information (alleviated by serial mode); biographical bias of human coders affects F1; API costs are relatively high, and long dreams may have formatting errors.
Section 07
Applicable to large-scale analysis, cross-cultural research, longitudinal tracking, and teaching training, helping researchers save coding time and focus on analysis and interpretation.
Section 08
llm_dream_coder does not replace researchers' judgments; it automates tedious coding. Future improvements to other modules will bring greater value to dream research.
Keep going with more reads from the same topic.
Splinter is a minimalist, high-performance key-value (KV) and vector storage system enabling zero-latency inter-process communication via shared memory and atomic operations. With only 766 lines of core code, it supports millions of operations per second and 768-dimensional vector storage, offering a new architectural approach for local LLM inference and data-intensive applications.
Folkering OS is the world's first AI-native bare-metal operating system, entirely written in Rust no_std without relying on Linux, POSIX, or libc. It can generate commands from scratch, compile them into WASM, and run them in 10 seconds, achieving true self-evolution.
This project explores how to use Large Language Models (LLMs) to detect logical vulnerabilities in Ethereum smart contracts, especially business logic flaws that are difficult for traditional static analysis tools to capture, such as economic manipulation and execution order errors.
This article introduces an open-source project for building modern large language models from scratch. The project evolves from a GPT-2-style basic architecture to a Llama 2/3-style production-level implementation, covering complete tutorial implementations of key technologies such as RMSNorm, RoPE, SwiGLU, GQA, and MoE.