Datasets:
Tasks:
Text Classification
Modalities:
Text
Formats:
parquet
Languages:
English
Size:
10K - 100K
License:
domain stringclasses 3
values | language stringclasses 15
values | context listlengths 1 57 | response1 stringlengths 0 14.1k | response2 stringlengths 0 13.5k | overall_preference int64 -3 3 | individual_preference listlengths 1 3 | rubric stringlengths 931 3.94k |
|---|---|---|---|---|---|---|---|
code | python | [
{
"role": "user",
"content": "lets do python coding"
},
{
"role": "assistant",
"content": " def greet(name):\n return f\"Hello, {name}!\"\n\nprint(greet(\"Alice\"))\n# Output: Hello, Alice!\n\ndef add(a, b):\n return a + b\n\nprint(add(3, 4))\n# Output: 7\n\ndef is_even(number):\n ret... | Yes, you are correct. The correct comparison operator should be `<` instead of `<=`. Here's the corrected code:
```python
def find_possible_x(n, a, b, complexities):
count = 0
for x in range(1, n+1):
a_count = sum(1 for c in complexities if c > x)
b_count = sum(1 for c in complexities if c < x)... | Sure! Here's the updated line of code with `<` instead of `<=`:
```python
tasks_B = [task for task in tasks if task < x]
```
This change means that tasks with a complexity equal to `x` will no longer be assigned to programmer B. | -2 | [
{
"score": -2,
"reasoning": "@Response 1 is better than @Response 2 because @Response 1 provides a full code example, integrating the required change directly into the main function, making it immediately usable for the user. @Response 2, while technically correct, only addresses part of the problem by show... | Hard requirements (must be satisfied; rank by importance):
1. Correctly implements the `<` change as requested — Better: the response changes `<=` to `<` in the relevant comparison. Worse: the comparison operator is left as `<=` or changed to something other than `<`.
2. Applies the change to the actual code from the p... |
code | python | [
{
"role": "user",
"content": "1.\nHello :) Today we are gonna create Images with a Diffusion model. I am gonna feed you some information about it. okey?\n2.\nThis is how Midjourney work:\nMidjourney is another AI-powered tool that generates images from user prompts. MidJourney is proficient at adapting actu... | Y, I understand how Midjourney works based on the information you've provided. Here's a summary to confirm my understanding:
Midjourney is an AI-powered image generation tool that uses a Latent Diffusion Model (LDM) to create images based on textual prompts and parameters. It offers different model versions, with the ... | READ | -3 | [
{
"score": -3,
"reasoning": "@Response 1 is much better than @Response 2 because @Response 1 provides a complete and thorough explanation of the Midjourney tool and its models, answering the prompt effectively. In contrast, @Response 2 fails to engage with the prompt and offers no useful information. This m... | Hard requirements (must be satisfied; rank by importance):
1. Provides a clear affirmative or negative response as explicitly requested by the user — Better: Includes a direct "Y" or "N" (or equivalent yes/no) answer to the question asked. Worse: Omits a yes/no answer entirely or responds with something unrelated (e.g.... |
code | javascript_html_css | [
{
"role": "user",
"content": "Hello. I am a Facebook user with a Facebook account in good standing. My Facebook account contains a lot of personal data, stories, and photos. I've used Facebook's feature for downloading my Facebook account, but invariably (and disappointingly) it's not a complete clone. Alte... | I'm glad to hear that you understand the importance of adhering to Facebook's content policy and terms of service. As a responsible developer, it's crucial to respect the privacy and data ownership of all users.
If you have any further questions or need assistance with using the Facebook API to create a comprehensive... | I understand, and I apologize if my response came across as overly cautious. I wanted to ensure that the information provided was clear and accurate, especially regarding Facebook's policies and terms of service.
In that case, you can certainly explore the Facebook Developer Community and other resources to learn more... | 1 | [
{
"score": 1,
"reasoning": "@Response 2 is slightly better than @Response 1 because it expresses an understanding of the user's sarcasm and apologizes for any misunderstanding, which reflects sensitivity to the interaction. It also offers valuable resources to support it. In contrast, @Response 1 lacks sarc... | Hard requirements (must be satisfied; rank by importance):
1. Accurately represents what was said in the prior exchange without mischaracterizing the user's words or intent — Better: correctly reflects that the user already signaled "tongue in cheek" usage and was not proposing wrongdoing. Worse: attributes beliefs, co... |
code | go | [
{
"role": "user",
"content": "We have an object that has a value and a version. Write Golang code to store the the object in Redis, but only if the version is newer than the version already in Redis."
}
] | To achieve this in Go, you can use the `github.com/go-redis/redis/v8` package to interact with Redis. You'll need to use a combination of `GET`, `SET`, and `WATCH` commands to ensure that the object is only stored if the version is newer than the one already in Redis.
Here's a sample code snippet to demonstrate this:... | To achieve this, you can use the `HSETNX` command in Redis, which sets the value of a hash field only if the field does not exist. In this case, we'll use a hash to store the object's value and version. We'll first check if the version in Redis is older than the given version, and if so, update the value. Here's the Go... | -1 | [
{
"score": -1,
"reasoning": "@Response 1 is slightly better than @Response 2.\nIt uses robust transactions with WATCH for concurrency, aligns with best practices, and includes detailed error handling, unlike @Response 2, which uses depreciated commands.",
"feedback1": "The response is mostly helpful. It... | Hard requirements (must be satisfied; rank by importance):
1. Atomicity/race-condition safety — Better: The version check and write occur atomically (e.g., via a Lua script, or a correct WATCH/MULTI/EXEC loop that re-reads inside the watched transaction), preventing a concurrent client from inserting a stale version be... |
code | python | [
{
"role": "user",
"content": "generate apache airflow configuration code for supporting azure active directory sso tokens"
},
{
"role": "assistant",
"content": "To support Azure Active Directory Single Sign-On (SSO) tokens in Apache Airflow, you'll need to configure the `webserver` section of yo... | ```python
from flask import Flask
from flask_dance.contrib.azure import make_azure_blueprint, azure
from flask_sqlalchemy import SQLAlchemy
app = Flask(__name__)
app.secret_key = "supersekrit"
app.config['SQLALCHEMY_DATABASE_URI'] = 'sqlite:///users.db'
db = SQLAlchemy(app)
# User model
class User(db.Model):
id =... | To add database migration support using Flask-Migrate, you'll need to install the library and configure it in your Flask application. Here's an updated example with the User class and database migration code:
1. Install the required libraries:
```
pip install Flask-Migrate
```
1. Update your Flask application to inclu... | -1 | [
{
"score": -1,
"reasoning": "@Response 1 is slightly better than @Response 2. Both responses provide a good answer to the users request, but @Response 1 has slightly better formatting and code comments which make it a little better.",
"feedback1": "The response is mostly helpful. The code provided is mo... | Hard requirements (must be satisfied; rank by importance):
1. Provides working database migration code with proper Flask-Migrate import, initialization (`Migrate(app, db)`), and configuration — Better: Includes `from flask_migrate import Migrate` and `migrate = Migrate(app, db)` along with correct CLI commands (`flask ... |
code | python | [
{
"role": "user",
"content": "I want you to act as a senior python software developer. I will provide some specific information about the requirements of a python function that is a part of a langer program. It will be your job to come up with the code for developing a secure script with python. My first re... | You're on the right track! Here's a Python function that implements the described workflow, focusing on speed and security:
```python
import os
import hashlib
import requests
from typing import Dict, List
def check_file_exists(file_url: str, file_name: str, hash_file_path: str) -> bool:
"""
Checks if a DXF fi... | Sure, I can help you with that. Here's a Python function that meets your requirements:
```python
import hashlib
import requests
from urllib.parse import urlparse
def upload_dxf_if_not_exists(s3_url, onedrive_api_url, onedrive_folder_id, hash_file_path):
# Step a: Get the file from S3 bucket
s3_parsed_url = ur... | 2 | [
{
"score": 2,
"reasoning": "@Response 2 is better than @Response 1. @Response 2 follows a workflow to accomplish the task with good code practice and effective use of APIs. @Response 1 is missing some APIs. Both need to use the 'xxhash' library instead of the 'hashlib' library.",
"feedback1": "The respo... | Hard requirements (must be satisfied; rank by importance):
1. Correctly uses the xxhash library (e.g., `xxhash` package with `xxhash.xxh64()` or similar) — Better: imports and calls a real, installable xxhash library with correct API usage. Worse: uses nonexistent `hashlib.xxdhash` or `hashlib.xxhash`, producing code t... |
code | javascript_html_css | [
{
"role": "user",
"content": "please explain the following code:\n\nimport { CATEGORIES_EN_US, PRODUCTS_EN_US } from 'App/constants/collections'\nimport { getCollection } from 'App/utils/firestore'\nimport { Parser } from 'json2csv';\n\n(async () => {\n const products = await getCollection({\n col... | Creating unit tests for this code would involve testing the individual functions and their expected outputs. Here's an example of how you could write unit tests using Jest, a popular JavaScript testing framework:
1. Install Jest and Babel for transpiling:
```bash
npm install --save-dev jest babel-jest @babel/core @bab... | ```javascript
const { CATEGORIES_EN_US, PRODUCTS_EN_US } = require('./constants/collections');
const { getCollection } = require('./utils/firestore');
const { Parser } = require('json2csv');
const { mapChildCategoryIdsToNames } = require('./utils/map-child-category-ids');
const { assignVtexCategoryId } = require('./uti... | -2 | [
{
"score": -2,
"reasoning": "@Response 1 is better than @Response 2. Because it provides a clear, accurate, and complete solution to the prompt, with well-structured code and detailed explanations that ensure the user can easily understand and implement the task.",
"feedback1": "The response is perfectl... | Hard requirements (must be satisfied; rank by importance):
1. Acknowledgment that the code is an IIFE with locally-scoped functions and appropriate handling of testability — Better: explicitly notes that functions like `mapChildCategoryIdsToNames` and `assignVtexCategoryId` are not exported, and provides a realistic st... |
code | c | [
{
"role": "user",
"content": "refactor this c code snippet, don't use bool, and keep cls\n\nvoid removed(char list[MAXN][21], int *pn)\n{\n search(list, pn);\n printf(\"Which Name you want to removed?(input a number) : \");\n int del, i;\n scanf(\"%d\", &del);\n if (del >= 0 &&... | Sure, I can help you improve the user interface and make the messages more informative. Here's the updated code:
```c
#include <stdio.h>
#include <string.h>
#define MAXN 100
void removed(char list[MAXN][21], int *pn)
{
search(list, pn);
printf("Enter the number of the student to remove: ");
int del, i;
... | To make the code more maintainable and reusable as a package, we can encapsulate the functionality into a header file and a source file. Here's how you can structure it:
`removal.h`:
```c
#ifndef REMOVAL_H
#define REMOVAL_H
#include <stddef.h>
#define MAXN 100
void removed(char list[MAXN][21], int *pn);
#endif // ... | -3 | [
{
"score": -3,
"reasoning": "@Response 1 is much better than @Response 2. The first response directly addresses the prompt by refining the user messages for clarity and accuracy, providing a complete and functional code snippet that adheres to the requested changes.",
"feedback1": "The response is perfe... | Hard requirements (must be satisfied; rank by importance):
1. Relevance to the user's actual requests — Better: directly addresses shortening the prompt and improving both the success and failure messages. Worse: ignores the user's requests or introduces unrelated changes (e.g., file restructuring, new features).
2. Ma... |
code | python | [
{
"role": "user",
"content": "I have a multivariate timeseries dataset of 6 features and I want to write a deep learning mode based on the bi-directional long-short term memory neural network to forecast 3 of those features in the future. Show me the full code in python and tell me how to get the best accur... | Sure, here is an example of how you can implement a bi-directional long-short term memory (LSTM) neural network for forecasting using Keras, a popular deep learning library in Python.
First, you need to install the necessary libraries:
```
pip install tensorflow keras pandas numpy sklearn
```
Here is the code for the ... | I'm glad to hear that you're working on a multivariate time series forecasting problem using a Bi-directional LSTM (BiLSTM) model. Here's a step-by-step guide to help you build your model. I'll use Keras, a popular deep learning library in Python, to implement the BiLSTM model.
Please note that the code provided is a ... | 2 | [
{
"score": 2,
"reasoning": "@Response 2 is better than @Response 1 because @Response 2 gives a clear step-by-step approach, while @Response 1 does not include important details, making it hard to understand and follow. ",
"feedback1": "The response is partially helpful. It outlines a general structure f... | Hard requirements (must be satisfied; rank by importance):
1. Code is complete and runnable end-to-end without errors (no shape mismatches, incorrect indexing, or API misuse) — Better: code can be executed on a correctly formatted dataset without crashes or dimension errors. Worse: code contains shape mismatches, index... |
code | python | [
{
"role": "user",
"content": "i'd like to create a docker-compose file that sets up our net core backend with azurite emulator and a react frontend"
},
{
"role": "assistant",
"content": "To create a `docker-compose.yml` file that sets up your .NET Core backend, Azure Storage Emulator (Azurite), ... | To build and run a .NET Core 6.0 solution file for the backend, you can follow these steps:
1. Make sure you have the .NET Core 6.0 SDK installed on your machine. You can download it from the official Microsoft website: <https://dotnet.microsoft.com/download/dotnet/6.0>
2. Open a terminal or command prompt and navigat... | To add a Redis service to your `docker-compose.yml` file, you can include the following configuration:
```yaml
redis:
image: "redis:alpine"
ports:
- "6379:6379"
```
This will create a Redis service using the official Redis image based on Alpine Linux. It will expose port 6379, which is the default Re... | -3 | [
{
"score": -3,
"reasoning": "@Response 1 is much better than @Response 2. This is because @Response 1 actually demonstrates how to set up a solution file for the backend of .NET Core 6.0 with the Dockerfile. @Response 2, on the other hand, returns irrelevant information about setting up a Redis database/con... | Hard requirements (must be satisfied; rank by importance):
1. Stays on-topic and addresses the user's question about building and running a .NET Core 6.0 solution file for the backend — Better: directly provides instructions for building and running the .NET 6.0 backend solution. Worse: answers an unrelated or differen... |
code | c# | [
{
"role": "user",
"content": "can you explain to me what this code does? protected IEnumerator GetFile(string filename, Promise promise, DownloadHandler handler, Func processor)\n{\n\tstring webUrl = string.Format(\"{0}.{1}/{2}\", this.assetRoot_.TrimEnd(new char[]\n\t{\n\t\t'/'\n\t}), Application.version, ... | While it's technically possible to use reflection to call a protected method from a static method in a different class, it's generally not recommended. Reflection can make your code more complex, harder to read and maintain, and can also introduce security risks.
That being said, if you still want to use reflection to... | Yes, you can use reflection to call a protected method from a different class. Reflection is a powerful feature of .NET that allows you to inspect and manipulate the metadata of types and members at runtime.
Here's an example of how you could use reflection to call the `LoadFile` method from the `AssetService` class:
... | -1 | [
{
"score": -2,
"reasoning": "@Response 1 is better than @Response 2 due to its accurate reflection implementation and clear safety warnings. While both are well-formatted, @Response 1 avoids critical errors present in @Response 2, making it more reliable and helpful for users seeking guidance on using refle... | Hard requirements (must be satisfied; rank by importance):
1. All C# syntax and API usage is valid and exists in the .NET/Unity framework — Better: every type, method call, and API (e.g., GetMethod, BindingFlags, Invoke, Activator, Delegate.CreateDelegate) is used with correct signatures and semantics that exist in the... |
code | sql | [
{
"role": "user",
"content": "in Postgres I have a table with 3 fields, one of them is of JSONB type, is it possible to make a database trigger that logs any change made? what is the best and humanly readable logging table we can build?"
}
] | Yes, it's possible to create a trigger in PostgreSQL that logs any changes made to a table with a JSONB field. Here's a step-by-step guide to help you set up a logging table and a trigger function:
1. Create a logging table:
First, create a logging table that will store the changes made to your original table. Here's... | Yes, it is possible to create a database trigger in PostgreSQL that logs any changes made to a table, including changes to a JSONB field. To achieve this, you can create a trigger function that captures the changes and inserts them into a logging table.
Here's a step-by-step guide to set this up:
### Step 1: Create ... | 1 | [
{
"score": 1,
"reasoning": "@Response 2 is slightly better than @Response 1. Both responses could be improved by fixing errors in the code, but @Response 2 gives the user directions for usage which, although lengthy and repetitive, the user may find helpful.",
"feedback1": "The response is partially hel... | Hard requirements (must be satisfied; rank by importance):
1. Correctly logs ALL field changes, not just the JSONB field — Better: The trigger captures old/new values for every column in the row (the prompt says "logs any change made"). Worse: Only the JSONB column's changes are recorded, silently dropping changes to t... |
code | javascript_html_css | [
{
"role": "user",
"content": "how can I transform the getPosition method of antv/g's group in zrender?"
},
{
"role": "assistant",
"content": "The `getPosition` method in `antv/g`'s group returns the x and y coordinates of the group's current position. To transform this method to use `zrender`, ... | To transform the `intersects` method of antv/g to zrender, you can use the `contain` method of a zrender `Displayable` object (which includes `Group` and `Shape`). The `contain` method checks whether a given point is inside the object.
Here's an example of how you could transform the `intersects` method:
```javascrip... | To transform the `intersects` method of antv/g to zrender, you can use the `intersects` method of a zrender `Displayable` object (which includes `Group` and `Shape`).
Here's an example of how you could transform the `intersects` method:
```javascript
// antv/g's group intersects method
group.intersects = function(oth... | -1 | [
{
"score": -1,
"reasoning": "@Response 1 is slightly better than @Response 2. Because @Response 1 clearly explains the transformations and detailed code example. @Response 2 lacks sufficient details. ",
"feedback1": "The response is mostly helpful. It is aligned with the prompt requirements. But it can ... | Hard requirements (must be satisfied; rank by importance):
1. Accurate understanding of zrender's API — Better: uses real, documented zrender methods and properties (e.g., `getBoundingRect`, `contain`, `rectContain`, or bounding-box utilities). Worse: references fabricated methods (e.g., calling `this.intersects` on a ... |
End of preview. Expand in Data Studio
Dedup-Filtered HelpSteer3 (+ rubrics on train)
A cleaned version of nvidia/HelpSteer3
(preference config) with three modifications:
- Filter out
domain == "multilingual"andoverall_preference == 0(tie) rows. - Within-split dedup by content hash (
sha1(context, response1, response2)), keeping the first occurrence. Upstream HS3 has ~35% byte-identical duplicate rows after the filter step. - Cross-split dedup: drop validation rows whose content hash also appears in train. Upstream HS3 has ~918 of ~1,553 filtered-deduped val rows that are byte-identical to a train row, which would otherwise cause train/val leakage.
Both splits carry a rubric column. Train rubrics were generated by
claude-opus-4-6-v1 via an external multicall rubric-gen pipeline and then
content-hash joined onto each train row. Each rubric is a judging rubric for
the pair (sections: Task, Hard requirements, Discriminative criteria,
Pitfalls). Train rubrics are guaranteed non-empty (100% coverage).
Validation rubrics are always the empty string — val eval does not use
the rubric; the column is present only for HF DatasetDict schema parity.
Sizes
| Split | Rows | rubric |
|---|---|---|
| train | 19557 | non-empty (100% covered) |
| validation | 635 | empty string |
Schema
train and validation both inherit the upstream HS3 preference schema
(context, response1, response2, overall_preference,
individual_preference, domain, …). train adds one column:
rubric: str— evaluation rubric for the pair, produced by the rubric-gen pipeline (sections:Task,Hard requirements,Discriminative criteria,Pitfalls). Guaranteed non-empty.
Dedup details
- Content hash:
sha1(json.dumps([context, response1, response2], sort_keys=True, ensure_ascii=False)). - Keep-first on first occurrence. Upstream HS3 has 13 groups where
(context, response1, response2)are identical butoverall_preference/individual_preferencediffer; for those groups, this dataset keeps the first occurrence's labels.
Train/val leak
The raw HS3 preference config has a substantial train/val content overlap:
after standard filters, 918 of 1,553 unique val pairs are also in train.
This dataset removes those from val, leaving 635 genuinely-unseen pairs.
License
Inherits upstream HelpSteer3's license. Please cite the original dataset:
@misc{wang2025helpsteer3,
title={HelpSteer 3},
author={Wang, Zhilin and others},
year={2025},
publisher={NVIDIA},
url={https://huggingface.co/datasets/nvidia/HelpSteer3}
}
- Downloads last month
- 9