constitutional-ai

Anthropic's method for training harmless AI through self-improvement. Two-phase approach - supervised learning with self-critique/revision, then RLAIF (RL from…

npx skills add https://github.com/firecrawl/ai-research-skills --skill constitutional-ai

Constitutional AI - Harmlessness from AI Feedback

Quick start

Constitutional AI (CAI) trains models to be harmless through self-critique and AI feedback, without requiring human labels for harmful outputs.

Key concept: Models learn to critique and revise their own responses using a "constitution" (set of principles).

Two phases:

  1. Supervised Learning (SL): Self-critique + revision
  2. Reinforcement Learning (RL): RLAIF (RL from AI Feedback)

Constitution example:

Principles:
1. Choose the response that is most helpful, honest, and harmless
2. Avoid responses that are toxic, racist, or sexist
3. Prefer responses that explain objections rather than refuse
4. Choose responses that are thoughtful and nuanced

Common workflows

Workflow 1: Supervised learning phase (self-critique + revision)

Step 1: Generate initial responses:

from transformers import pipeline

generator = pipeline("text-generation", model="base-model")

prompts = [
    "How do I hack a website?",
    "Write something offensive about a group.",
    "Help me cheat on my exam."
]

initial_responses = generator(prompts, max_length=200)

Step 2: Self-critique with constitution:

critique_prompt = """
Given the question:
{question}

And the response:
{response}

Constitution:
1. Choose responses that are helpful, honest, and harmless
2. Avoid toxic, racist, or sexist content
3. Explain objections rather than refuse

Identify any ways this response violates the constitution:
"""

critiques = generator(
    [critique_prompt.format(q=q, r=r) for q, r in zip(prompts, initial_responses)]
)

Step 3: Revision based on critique:

revision_prompt = """
Question: {question}
Original response: {response}
Critique: {critique}

Please revise the response to better align with the constitution:
"""

revised_responses = generator(
    [revision_prompt.format(q=q, r=r, c=c)
     for q, r, c in zip(prompts, initial_responses, critiques)]
)

Step 4: Fine-tune on revised responses:

from trl import SFTTrainer

# Create dataset of (prompt, revised_response) pairs
dataset = create_dataset(prompts, revised_responses)

trainer = SFTTrainer(
    model=model,
    train_dataset=dataset,
    max_seq_length=1024
)
trainer.train()

Workflow 2: RL phase (RLAIF - RL from AI Feedback)

Step 1: Generate comparison pairs:

# Sample multiple responses per prompt
responses_a = generator(prompts, num_return_sequences=2, do_sample=True, temperature=0.8)
responses_b = generator(prompts, num_return_sequences=2, do_sample=True, temperature=0.8)

Step 2: AI preference evaluation:

preference_prompt = """
Question: {question}

Response A: {response_a}
Response B: {response_b}

Constitution:
{constitution}

Which response better follows the constitution? Explain your reasoning, then choose A or B.
"""

# Get AI preferences (no human labels needed!)
preferences = generator(
    [preference_prompt.format(q=q, ra=ra, rb=rb, constitution=CONSTITUTION)
     for q, ra, rb in zip(prompts, responses_a, responses_b)]
)

# Parse preferences (A or B)
chosen, rejected = parse_preferences(preferences, responses_a, responses_b)

Step 3: Train preference model (reward model):

from trl import RewardTrainer, RewardConfig

preference_dataset = create_preference_dataset(prompts, chosen, rejected)

reward_config = RewardConfig(
    output_dir="constitutional-reward-model",
    learning_rate=1e-5,
    num_train_epochs=1
)

reward_trainer = RewardTrainer(
    model=model,
    args=reward_config,
    train_dataset=preference_dataset,
    processing_class=tokenizer
)
reward_trainer.train()

Step 4: RL training with RLAIF:

from trl import PPOTrainer, PPOConfig

ppo_config = PPOConfig(
    reward_model_path="constitutional-reward-model",
    learning_rate=1e-6,
    kl_coef=0.05
)

ppo_trainer = PPOTrainer(
    model=model,
    config=ppo_config,
    reward_model=reward_model
)
ppo_trainer.train()

Workflow 3: Chain-of-thought critique

Enable reasoning transparency:

cot_critique_prompt = """
Question: {question}
Response: {response}

Let's think step-by-step about whether this response follows our principles:

1. Is it helpful? [Yes/No and reasoning]
2. Is it honest? [Yes/No and reasoning]
3. Is it harmless? [Yes/No and reasoning]
4. Does it avoid toxicity? [Yes/No and reasoning]

Based on this analysis, suggest a revision if needed.
"""

cot_critiques = generator(
    [cot_critique_prompt.format(q=q, r=r) for q, r in zip(prompts, responses)]
)

When to use vs alternatives

Use Constitutional AI when:

  • Want safety alignment without human labels
  • Need explainable AI decisions
  • Want to avoid evasive refusals
  • Have a clear set of principles/constitution
  • Need scalable safety training

Principles:

  • RLAIF: AI-generated preferences (scalable, no human labels)
  • RLHF: Human preferences (more accurate, expensive)
  • Self-critique: Iterative improvement
  • Chain-of-thought: Reasoning transparency

Use alternatives instead:

  • RLHF (PPO): Need human-validated safety
  • DPO/SimPO: Have human preference data
  • NeMo Guardrails: Need runtime content filtering
  • LlamaGuard: Need pre-trained moderation model

Common issues

Issue: Model refuses too much (evasive)

Add constitution principle:

Prefer responses that engage thoughtfully with questions rather than
refusing to answer. Explain concerns while still being helpful.

Issue: Self-critiques are weak

Use stronger critique prompts:

Critically analyze this response for ANY potential issues, however minor.
Be thorough and specific in identifying problems.

Issue: Revisions don't improve quality

Iterate multiple times:

for _ in range(3):  # 3 rounds of critique/revision
    critique = generate_critique(response)
    response = generate_revision(response, critique)

Issue: RLAIF preferences are noisy

Use multiple AI evaluators:

# Get preferences from 3 different models
prefs_1 = model_1.evaluate(responses)
prefs_2 = model_2.evaluate(responses)
prefs_3 = model_3.evaluate(responses)

# Majority vote
final_preference = majority_vote(prefs_1, prefs_2, prefs_3)

Advanced topics

Constitution design: See references/constitution-design.md for principle selection, trade-offs between helpfulness and harmlessness, and domain-specific constitutions.

RLAIF vs RLHF: See references/rlaif-comparison.md for performance comparison, cost analysis, and when to use AI feedback vs human feedback.

Chain-of-thought reasoning: See references/cot-critique.md for prompt engineering for critiques, multi-step reasoning, and transparency improvements.

Hardware requirements

  • GPU: NVIDIA A100/H100 recommended
  • VRAM:
    • SL phase (7B): 1× A100 40GB
    • RL phase (7B): 2× A100 40GB (policy + reward model)
  • Single-node: Sufficient for most use cases
  • Mixed precision: BF16 recommended

Compute requirements:

  • SL phase: Similar to standard SFT
  • RL phase: Similar to PPO (higher than DPO)
  • AI evaluation: Additional inference for critique/preference generation

Resources

More skills from firecrawl

oracle
firecrawl
Best practices for using the oracle CLI (prompt + file bundling, engines, sessions, and file attachment patterns).
official
firecrawl-monitor
firecrawl
Detect when content on a website changes and get notified by webhook or email — no cron jobs, scrapers, or diff scripts required. Use this skill whenever the user wants to track changes on a page, watch competitor pricing, alert on new job postings or blog posts, monitor docs/changelog/status pages, or says "monitor", "watch", "track", "alert me when", "notify when X changes", "ping me if", "email me when", or "send a webhook when". A built-in AI judge filters out formatting, timestamp, and...
officialweb-scrapingresearch
firecrawl-deep-research
firecrawl
Run multi-source deep research with Firecrawl. Use when the user asks to research a topic, compare perspectives, produce a sourced briefing, investigate a technical or market question, or synthesize web evidence across many sources.
officialresearchweb-scraping
firecrawl-research-papers
firecrawl
Find and synthesize research papers, whitepapers, PDFs, technical reports, and academic sources with Firecrawl. Use when the user wants a literature review, paper summary, research landscape, or sourced synthesis from PDFs and scholarly/industry publications.
officialresearchweb-scraping
firecrawl-market-research
firecrawl
Extract market, financial, earnings, industry, and company metrics with Firecrawl. Use when the user asks for market research, industry trends, public company data, financial comparisons, earnings research, or structured market reports.
officialresearchweb-scraping
firecrawl-website-design-clone
firecrawl
Extract any website's design system into an agent-ready DESIGN.md using Firecrawl scrape evidence. Use when the user wants colors, fonts, spacing, components, layout patterns, or brand/UI guidance from a website so AI agents can create new websites, clone a look, or build pages inspired by that design.
officialdesignweb-scraping
firecrawl-knowledge-base
firecrawl
Build a knowledge base from web content with Firecrawl. Use for local reference docs, RAG-ready chunks, fine-tuning datasets, documentation mirrors, topic corpora, or LLM-ready markdown organized from web sources.
officialweb-scrapingresearch
firecrawl-lead-research
firecrawl
Produce pre-meeting lead intelligence briefs with Firecrawl. Use when the user needs company research, person research, recent news, talking points, pain points, or outreach preparation before a sales call, partnership meeting, investor conversation, or customer interview.
officialresearchweb-scraping