Bible Network Crypto DeFi Onchain RWA AI Agent Stablecoin Chain SAFU CryptoTax DeFAI AGI Claude Me Claude Skill Claude Design Claude Cowork
Independent Media
Not affiliated with any project
Exploring the Frontier of AI Intelligence
claude-me.com
LATEST
Eight Principles Straight From Anthropic: If Claude Keeps Getting Worse, the Problem Is Probably How You're Using It  ·  Claude Code Can Now 'Loop' Through Work on Its Own: Anthropic Releases Four Loop Mode Guide That Lets AI Run Entire Processes  ·  Claude Cowork Honest Review: Three Months Later — What Actually Saved Time, What Made Me Regret Automating It  ·  Two Major Stories to Open July: Claude Sonnet 5 Launches, Fable 5 Export Controls Lifted and Globally Restored  ·  What Is RAG: Why Claude Can't Read Your Company Intranet — And How to Fix That  ·  Claude Cowork Advanced Workflows: From 'Hand Off One Task' to 'Run an Entire Process' — Three Real-World Templates
fundamentals

How Claude Learns to Be "Helpful to Humans": RLHF and Constitutional AI Explained

30-Second Version · For the impatient
RLHF teaches Claude what responses humans prefer. Constitutional AI teaches it what responses are actually right. Their combination is what makes Claude both helpful and honest.

Full Explanation +
01 · Why did this happen?

Claude's training involves two main stages: pre-training (learning language patterns) and Alignment training (learning to be "helpful to humans"). The main alignment training methods are RLHF (guiding the model with human feedback preferences) and Constitutional AI (self-evaluation based on a set of explicit behavioral principles). Their combination enables Claude to generate useful responses while honestly acknowledging uncertainty and providing meaningful explanations when declining requests.

02 · What is the mechanism?

RLHF was systematized by OpenAI in the 2017-early 2020s and applied at scale in InstructGPT training, later becoming the core method for training ChatGPT. Anthropic's Constitutional AI is an innovation building on RLHF, addressing the problem of RLHF's dependence on human preference annotation — annotator biases and inconsistencies directly affect trained model behavior. Constitutional AI attempts to replace subjective preference judgments with explicit principles.

03 · How does it affect me?

Understanding RLHF and Constitutional AI helps explain behavioral differences between Claude and other AI tools. Purely RLHF-trained systems are prone to "sycophancy" — tending to tell users what they want to hear rather than truthful answers. Constitutional AI's addition makes Claude notably different on this point: it's trained to remain honest even when users don't like the answer, which explains why Claude sometimes gives responses that differ from your expectations rather than simply echoing your viewpoint.

04 · What should I do?

Translate understanding of RLHF and Constitutional AI into practical usage techniques: if you want honest feedback rather than flattery, explicitly tell Claude "I don't need you to agree with my viewpoint — I need you to tell me where the problems are"; if Claude declines your request, asking "why?" typically gets a meaningful explanation rather than a formulaic "I can't help with that"; if you're unsure whether Claude's answer is accurate, directly ask "how confident are you in this answer? Are there parts you're uncertain about?" — it's trained to honestly express uncertainty in these situations.

Full Content +

A freshly trained language model is like a scholar who has read a vast amount of books but has no idea what "humans want." It can generate text — but not necessarily helpful, safe, or honest text. So how did Anthropic turn it into Claude?

Pre-training Is Just the Starting Point

All large language models begin with "pre-training" — exposing the model to massive text data to learn statistical patterns of language: what words typically follow what words, which sentence structures are grammatically valid, what characteristics different topics' text has.

After pre-training, the model's capabilities are already impressive — it can generate fluent text and understand complex semantics. But "generating fluent text" doesn't equal "being helpful to users." A freshly pre-trained model given a question might output any text related to that question, including incorrect information, biased viewpoints, or even harmful content — because all of these appear in its training data.

This is why a second stage is needed: Alignment Training.

RLHF: Guiding the Model With Human Preferences

RLHF (Reinforcement Learning from Human Feedback) is currently the most mainstream alignment training method — OpenAI used it to train ChatGPT, and Anthropic uses it too.

RLHF's process roughly works as follows:

Step 1: Supervised Fine-Tuning (SFT) Starting from the pre-trained model, human annotators demonstrate what ideal responses look like — for various types of questions, humans write what they consider the best responses, then this "demonstration data" is used to fine-tune the model.

Step 2: Train a Reward Model Human annotators evaluate and rank multiple different responses to the same question. These preference data train a "reward model" — whose job is to predict "how good would humans consider this response" and give it a score.

Step 3: Reinforcement Learning Optimization Using the reward model as a "teacher," the language model tries various different responses, receiving rewards for directions that score highly and penalties for those that score low. Through large volumes of "try-reward-adjust" cycles, the model gradually learns to generate responses humans consider better.

RLHF's Limitation: Human Annotators Are Biased

RLHF's core problem: it uses "human preference" as a proxy for "correct answers" — but human preference doesn't equal correct or safe. Annotators may prefer responses that sound more confident (even if inaccurate), prefer longer more detailed responses (even if shorter is better), or have systematic biases on cultural and political topics.

This is why Anthropic developed Constitutional AI in addition to RLHF.

Constitutional AI: Giving Claude a Clear Set of Value Principles

Constitutional AI is an alignment training approach Anthropic proposed in 2022. Its core idea: rather than having human annotators judge which responses are better, give the model an explicit set of behavioral principles and let it self-evaluate based on those principles.

This "constitution" is an explicitly written list of principles, such as: "should not help humans do things that could cause large-scale harm," "should be honest and not deceive users," "should respect human autonomy and not push particular viewpoints."

Step 1: AI Critique — let the model critique its own responses, identifying problems according to constitutional principles.

Step 2: AI Revision — let the model revise responses based on its own critique.

Step 3: AI Preference Labeling — Anthropic's key Constitutional AI innovation: let the model itself (not human annotators) score and rank responses, based on constitutional principles rather than purely subjective preferences.

The Combination of These Methods Shapes Claude's Personality

Claude is trained through a combination of RLHF and Constitutional AI, which explains several of its distinctive characteristics:

Honesty over flattery: Claude is trained to remain honest even when responses make users uncomfortable, rather than saying what they want to hear. Pure RLHF systems easily learn that "humans like being told what they want to hear" and trend toward sycophancy.

Proactively acknowledging limitations: Claude is trained to express uncertainty when uncertain, and acknowledge when things are beyond its capabilities — rather than pretending omniscience.

Refusals with reasoning: Claude doesn't decline requests because of a "prohibited items list" — it learned a set of principles and judges based on them. This lets it provide meaningful explanations for why it declines.

Diagram
RLHF + Constitutional AI 訓練流程對比流程圖對比 RLHF 和 Constitutional AI 兩種對齊訓練方法的運作方式:RLHF 依賴人類標記者的偏好評分,Constitutional AI 用明確的原則讓模型自我評估,說明兩者各自的優勢和限制,以及 Anthropic 如何結合兩者訓練 Claude。RLHF vs Constitutional AI — How Claude Learns ValuesRLHFHuman feedback drives alignment① Human annotators write ideal responses(Supervised Fine-Tuning)② Humans rank responses → Reward Model(which response is better?)③ RL optimizes toward high reward scores(model learns what humans prefer)⚠ Limitation: annotator biasesHumans may prefer confident-sounding butwrong answers → sycophancy riskConstitutional AIPrinciples drive self-evaluation① Model generates a response(same as before)② Model critiques itself using the Constitution("does this violate principle X?")③ Model revises → AI ranks responses(no human annotator needed for ranking)✓ Advantage: explicit, consistent principlesLess annotator bias · Scales betterClaude can explain WHY it declinesClaude Me · claude-me.com
Feel free to share. Please credit the source.
Ask a Question
Please enter at least 10 characters
Related Articles
Constitutional AI: How Anthropic Trains Values Into Claude
fundamentals · Jun 16
How Training Shapes Claude's Personality: The Complete Path From Pre-training to RLHF to Constitutional AI
fundamentals · Jun 05
How an LLM Actually Generates Text: A Real Explanation for Non-Engineers
fundamentals · Jun 17
Mechanistic Interpretability: Why Anthropic is Dissecting Claude's 'Brain' — Frontier AI Explainability Research
fundamentals · Jun 11
Related News
More Related Topics