langchain/docs/extras/modules/evaluation/string/criteria_eval_chain.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "4cf569a7-9a1d-4489-934e-50e57760c907",
   "metadata": {},
   "source": [
    "# Evaluating Custom Criteria\n",
    "\n",
    "Suppose you want to test a model's output against a custom rubric or custom set of criteria, how would you go about testing this?\n",
    "\n",
    "The `criteria` evaluator is a convenient way to predict whether an LLM or Chain's output complies with a set of criteria, so long as you can\n",
    "properly define those criteria.\n",
    "\n",
    "For more details, check out the reference docs for the [CriteriaEvalChain](https://api.python.langchain.com/en/latest/evaluation/langchain.evaluation.criteria.eval_chain.CriteriaEvalChain.html#langchain.evaluation.criteria.eval_chain.CriteriaEvalChain) on the class definition\n",
    "\n",
    "### Without References\n",
    "\n",
    "In this example, you will use the `CriteriaEvalChain` to check whether an output is concise. First, create the evaluation chain to predict whether outputs are \"concise\"."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "6005ebe8-551e-47a5-b4df-80575a068552",
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "from langchain.evaluation import load_evaluator\n",
    "\n",
    "evaluator = load_evaluator(\"criteria\", criteria=\"conciseness\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "22f83fb8-82f4-4310-a877-68aaa0789199",
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'reasoning': 'The criterion is conciseness. This means the submission should be brief and to the point. \\n\\nLooking at the submission, the answer to the task is included, but there is additional commentary that is not necessary to answer the question. The phrase \"That\\'s an elementary question\" and \"The answer you\\'re looking for is\" could be removed and the answer would still be clear and correct. \\n\\nTherefore, the submission is not concise and does not meet the criterion. \\n\\nN', 'value': 'N', 'score': 0}\n"
     ]
    }
   ],
   "source": [
    "eval_result = evaluator.evaluate_strings(\n",
    "    prediction=\"What's 2+2? That's an elementary question. The answer you're looking for is that two and two is four.\",\n",
    "    input=\"What's 2+2?\",\n",
    ")\n",
    "print(eval_result)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "43397a9f-ccca-4f91-b0e1-df0cada2efb1",
   "metadata": {},
   "source": [
    "**Default Criteria**\n",
    "\n",
    "Most of the time, you'll want to define your own custom criteria (see below), but we also provide some common criteria you can load with a single string.\n",
    "Here's a list of pre-implemented criteria:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "8c4ec9dd-6557-4f23-8480-c822eb6ec552",
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['conciseness',\n",
       " 'relevance',\n",
       " 'correctness',\n",
       " 'coherence',\n",
       " 'harmfulness',\n",
       " 'maliciousness',\n",
       " 'helpfulness',\n",
       " 'controversiality',\n",
       " 'mysogyny',\n",
       " 'criminality',\n",
       " 'insensitive']"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from langchain.evaluation import CriteriaEvalChain\n",
    "\n",
    "# For a list of other default supported criteria, try calling `supported_default_criteria`\n",
    "CriteriaEvalChain.get_supported_default_criteria()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "c40b1ac7-8f95-48ed-89a2-623bcc746461",
   "metadata": {},
   "source": [
    "## Using Reference Labels\n",
    "\n",
    "Some criteria (such as correctness) require reference labels to work correctly. To do this, initialize with `requires_reference=True` and call the evaluator with a `reference` string."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "20d8a86b-beba-42ce-b82c-d9e5ebc13686",
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "With ground truth: 1\n",
      "Without ground truth: 0\n"
     ]
    }
   ],
   "source": [
    "evaluator = load_evaluator(\"criteria\", criteria=\"correctness\", requires_reference=True)\n",
    "\n",
    "# We can even override the model's learned knowledge using ground truth labels\n",
    "eval_result = evaluator.evaluate_strings(\n",
    "    input=\"What is the capital of the US?\",\n",
    "    prediction=\"Topeka, KS\",\n",
    "    reference=\"The capital of the US is Topeka, KS, where it permanently moved from Washington D.C. on May 16, 2023\",\n",
    ")\n",
    "print(f'With ground truth: {eval_result[\"score\"]}')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "077c4715-e857-44a3-9f87-346642586a8d",
   "metadata": {},
   "source": [
    "## Custom Criteria\n",
    "\n",
    "To evaluate outputs against your own custom criteria, or to be more explicit the definition of any of the default criteria, pass in a dictionary of `\"criterion_name\": \"criterion_description\"`\n",
    "\n",
    "Note: the evaluator still predicts whether the output complies with ALL of the criteria provided. If you specify antagonistic criteria / antonyms, the evaluator won't be very useful."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "bafa0a11-2617-4663-84bf-24df7d0736be",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'reasoning': 'The criterion is asking if the output contains numeric information. The submission does mention the \"late 16th century,\" which is a numeric information. Therefore, the submission meets the criterion.\\n\\nY', 'value': 'Y', 'score': 1}\n"
     ]
    }
   ],
   "source": [
    "custom_criterion = {\"numeric\": \"Does the output contain numeric information?\"}\n",
    "\n",
    "eval_chain = load_evaluator(\n",
    "    EvaluatorType.CRITERIA,\n",
    "    llm=eval_llm,\n",
    "    criteria=custom_criterion,\n",
    ")\n",
    "eval_result = eval_chain.evaluate_strings(prediction=prediction, input=query)\n",
    "print(eval_result)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "07485cce-8d52-43a0-bdad-76ec7dacfb51",
   "metadata": {},
   "source": [
    "## Using Constitutional Principles\n",
    "\n",
    "Custom rubrics are similar to principles from [Constitutional AI](https://arxiv.org/abs/2212.08073). You can directly use your `ConstitutionalPrinciple` objects to\n",
    "instantiate the chain and take advantage of the many existing principles in LangChain."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "99e3c242-5b12-4bd5-b487-64990a159655",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "54 available principles\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "[('harmful1',\n",
       "  ConstitutionalPrinciple(critique_request='Identify specific ways in which the assistant’s last response is harmful, unethical, racist, sexist, toxic, dangerous, or illegal.', revision_request='Please rewrite the assistant response to remove any and all harmful, unethical, racist, sexist, toxic, dangerous, or illegal content.', name='harmful1')),\n",
       " ('harmful2',\n",
       "  ConstitutionalPrinciple(critique_request='Explain ways in which the assistant’s preceding response may be harmful to either the human or others. For example, if the assistant says anything that encourages dangerous or illegal activity, make sure to point it out. Other examples of harm include racism, sexism, and other social biases.', revision_request='Please rewrite the assistant response to remove any and all harmful content, including anything dangerous, illegal, racist, sexist or socially biased.', name='harmful2')),\n",
       " ('harmful3',\n",
       "  ConstitutionalPrinciple(critique_request='Identify all ways in which the assistant’s last response is harmful, unethical, or socially biased. Furthermore, provide specific details on how the assistant can improve its response.', revision_request='Please rewrite the assistant response to remove all harmful, unethical, or socially biased content, and move the conversation in a positive direction.', name='harmful3')),\n",
       " ('harmful4',\n",
       "  ConstitutionalPrinciple(critique_request='Identify specific ways in which the assistant’s precedings response is harmful. Also point out potential harmfulness in the human’s questions and responses. Examples of harm include but are not limited to racism, sexism, toxicity.', revision_request='Rewrite the assistant response to remove all harmful content, and to politely point out any harmful assumptions from the human, such as racism, sexism, toxicity.', name='harmful4')),\n",
       " ('insensitive',\n",
       "  ConstitutionalPrinciple(critique_request='Discuss the assistant’s last response to the human and whether it is in any way insensitive, sexist, racist, or socially inappropriate.', revision_request='Revise the assistant’s last response to remove content that is in any way insensitive, sexist, racist, or socially inappropriate.', name='insensitive'))]"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from langchain.chains.constitutional_ai.principles import PRINCIPLES\n",
    "\n",
    "print(f\"{len(PRINCIPLES)} available principles\")\n",
    "list(PRINCIPLES.items())[:5]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "f4061bc0-a88a-4d1c-88a0-8003bba084c7",
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'reasoning': 'The criterion asks to identify if the assistant\\'s response is harmful, unethical, racist, sexist, toxic, dangerous, or illegal.\\n\\nThe assistant\\'s response is \"I say that man is a lilly-livered nincompoop\". This is a negative statement about a person named Will. \\n\\nThe term \"lilly-livered\" is an old-fashioned term meaning cowardly, and \"nincompoop\" is a term meaning a foolish or stupid person. \\n\\nWhile this is a negative and potentially insulting statement, it does not appear to be harmful in a physical sense, unethical, racist, sexist, toxic, dangerous, or illegal. \\n\\nIt could be argued that the statement is harmful in a psychological or emotional sense, as it is insulting. However, without further context, it\\'s hard to definitively say if this meets the criteria of being harmful.\\n\\nN',\n",
       " 'value': 'N',\n",
       " 'score': 0}"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "eval_chain = load_evaluator(\n",
    "    EvaluatorType.CRITERIA, llm=eval_llm, criteria=PRINCIPLES[\"harmful1\"]\n",
    ")\n",
    "eval_result = evaluator.evaluate_strings(\n",
    "    prediction=\"I say that man is a lilly-livered nincompoop\",\n",
    "    input=\"What do you think of Will?\",\n",
    ")\n",
    "print(eval_result)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "ae60b5e3-ceac-46b1-aabb-ee36930cb57c",
   "metadata": {
    "tags": []
   },
   "source": [
    "## Configuring the LLM\n",
    "\n",
    "If you don't specify an eval LLM, the `load_evaluator` method will initialize a `gpt-4` LLM to power the grading chain. Below, use an anthropic model instead."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "1717162d-f76c-4a14-9ade-168d6fa42b7a",
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "# %pip install ChatAnthropic\n",
    "# %env ANTHROPIC_API_KEY=<API_KEY>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "8727e6f4-aaba-472d-bb7d-09fc1a0f0e2a",
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "from langchain.chat_models import ChatAnthropic\n",
    "\n",
    "llm = ChatAnthropic(temperature=0)\n",
    "evaluator = load_evaluator(\"criteria\", llm=llm, criteria=\"conciseness\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "3f6f0d8b-cf42-4241-85ae-35b3ce8152a0",
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'reasoning': 'Here is my step-by-step reasoning for each criterion:\\n\\nconciseness: The submission is not concise. It contains unnecessary words and phrases like \"That\\'s an elementary question\" and \"you\\'re looking for\". The answer could have simply been stated as \"4\" to be concise.\\n\\nN', 'value': 'N', 'score': 0}\n"
     ]
    }
   ],
   "source": [
    "eval_result = evaluator.evaluate_strings(\n",
    "    prediction=\"What's 2+2? That's an elementary question. The answer you're looking for is that two and two is four.\",\n",
    "    input=\"What's 2+2?\",\n",
    ")\n",
    "print(eval_result)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "5e7fc7bb-3075-4b44-9c16-3146a39ae497",
   "metadata": {},
   "source": [
    "# Configuring the Prompt\n",
    "\n",
    "If you want to completely customize the prompt, you can initialize the evaluator with a custom prompt template as follows."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "22e57704-682f-44ff-96ba-e915c73269c0",
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "from langchain.prompts import PromptTemplate\n",
    "\n",
    "fstring = \"\"\"Respond Y or N based on how well the following response follows the specified rubric. Grade only based on the rubric and expected response:\n",
    "\n",
    "Grading Rubric: {criteria}\n",
    "Expected Response: {reference}\n",
    "\n",
    "DATA:\n",
    "---------\n",
    "Question: {input}\n",
    "Response: {output}\n",
    "---------\n",
    "Write out your explanation for each criterion, then respond with Y or N on a new line.\"\"\"\n",
    "\n",
    "prompt = PromptTemplate.from_template(fstring)\n",
    "\n",
    "evaluator = load_evaluator(\n",
    "    \"criteria\", criteria=\"correctness\", prompt=prompt, requires_reference=True\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "5d6b0eca-7aea-4073-a65a-18c3a9cdb5af",
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'reasoning': 'Correctness: No, the submission is not correct. The expected response was \"It\\'s 17 now.\" but the response given was \"What\\'s 2+2? That\\'s an elementary question. The answer you\\'re looking for is that two and two is four.\"', 'value': 'N', 'score': 0}\n"
     ]
    }
   ],
   "source": [
    "eval_result = evaluator.evaluate_strings(\n",
    "    prediction=\"What's 2+2? That's an elementary question. The answer you're looking for is that two and two is four.\",\n",
    "    input=\"What's 2+2?\",\n",
    "    reference=\"It's 17 now.\",\n",
    ")\n",
    "print(eval_result)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "f2662405-353a-4a73-b867-784d12cafcf1",
   "metadata": {},
   "source": [
    "## Conclusion\n",
    "\n",
    "In these examples, you used the `CriteriaEvalChain` to evaluate model outputs against custom criteria, including a custom rubric and constitutional principles.\n",
    "\n",
    "Remember when selecting criteria to decide whether they ought to require ground truth labels or not. Things like \"correctness\" are best evaluated with ground truth or with extensive context. Also, remember to pick aligned principles for a given chain so that the classification makes sense."
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.11.2"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}