SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2402.02423 · ICLR · 2024

Uni-RLHF: Universal Platform and Benchmark Suite for Reinforcement Learning with Diverse Human Feedback

Jianye Hao, Yi Ma, Yan Zheng, Yifu Yuan, Jinyi Liu, Kai Zhao, Zibin Dong, Hebin Liang, Zhixin Feng

arXiv · PDF · Open in the Atlas

Code that ran

9 of 9 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (0 of them constructed an object rather than computing a result; 1 honoured the contract we drafted, 1 violated it and 3 have no contract verdict; 4 failures of our instrument; 0 no recorded run); 1 of them was also run on the host, not in the jail, by the peer-agreement lane that set its level, measured 2026-11-22; repository links from the Papers with Code archive attributed under CC BY-SA, no affiliation; bodies inline only under a permissive licence as GitHub detects it. Not a reproduction of the paper's results.

"Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results. This record, dated, as JSON: /api/ran/2402.02423.json · badge: /api/ran/2402.02423.svg

Each repository's role is a label: a classifier's output recorded on the link, or, where no role is recorded, what the Papers with Code archive marked. The line under it names its basis. Neither says who wrote the code.

RepositoryRoleRan
pickxiguapi/uni-rlhf-platform community (classifier label)
classifier basis not recorded on the link
4 ran · 0 no recorded run
pickxiguapi/clean-offline-rlhf reimplementation (classifier label)
classifier basis: classifier_2025
3 ran · 0 no recorded run
copy not recorded community
not in the paper's declared repositories
2 ran · 0 no recorded run
FunctionStatusWhere it lives
compute_mean_std Ran
ran; our contract draft was wrong, not the code
this paper's copy was not recorded; identical code first harvested from YangRui2015/RIQL
pointer only; the stored copy get_code serves, from YangRui2015/RIQL, is pointer only (licence: NONE) · get_code("639b4986c270df90")
How it ran

The body of this sample is withheld: this paper's copy of the code was not recorded (its link names no repository or source path), so no licence was read for it. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

allowed_file Ran
ran, violated the contract we drafted
pickxiguapi/uni-rlhf-platform/uni_rlhf/utils.py
code served (permissive licence) · get_code("2e291147f74aebf2")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned bool, ndim 0, dtype family float, shape () as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import numpy as np

ARGS = {
    'filename': 'test_data.hdf5'
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type bool, ndim 0, dtype family float, shape () — shape () as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
datetime_converter Ran
ran on a synthesized input
pickxiguapi/uni-rlhf-platform/uni_rlhf/utils.py
code served (permissive licence) · get_code("8cce8bc485699c38")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned str. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import datetime

def datetime_converter(o):
    if isinstance(o, datetime.datetime):
        return o.__str__()

# Create a datetime object to pass as the argument
dt_object = datetime.datetime(2024, 1, 15, 10, 30, 45)

ARGS = {
    "o": dt_object
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type str — a str output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a str output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
get_atari_dataset Ran
ran; our contract draft was wrong, not the code
pickxiguapi/clean-offline-rlhf/rlhf/train_reward_model.py
code served (permissive licence) · get_code("1b2a79e9b280b172")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned dict. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np

class MockAtariEnv:
    def get_dataset(self):
        return {
            'rewards': np.array([1.0, 2.0, 3.0, 4.0, 5.0]),
            'observations': np.random.randn(5, 84, 84, 4),
            'actions': np.array([0, 1, 2, 1, 0]),
            'dones': np.array([False, False, True, False, False]),
            'next_observations': np.random.randn(5, 84, 84, 4)
        }

env = MockAtariEnv()

ARGS = {
    'env': env
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type dict — a dict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dict output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
get_d4rl_dataset Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
pickxiguapi/clean-offline-rlhf/rlhf/train_reward_model.py
code served (permissive licence) · get_code("b18b08658373ee43")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned dict. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np
import gym

class MockEnv:
    """Mock environment that returns a dataset when d4rl.qlearning_dataset is called"""
    def __init__(self):
        self.observation_space = gym.spaces.Box(low=-1, high=1, shape=(4,), dtype=np.float32)
        self.action_space = gym.spaces.Box(low=-1, high=1, shape=(3,), dtype=np.float32)

class MockD4RL:
    """Mock d4rl module"""
    @staticmethod
    def qlearning_dataset(env):
        n_samples = 2
        return {
            'observations': np.random.randn(n_samples, 4).astype(np.float32),
            'actions': np.random.randn(n_samples, 3).astype(np.float32),
            'next_observations': np.random.randn(n_samples, 4).astype(np.float32),
            'rewards': np.random.randn(n_samples).astype(np.float32),
            'terminals': np.array([0, 1], dtype=bool)
        }

# Mock the d4rl module
import sys
mock_d4rl = MockD4RL()
sys.modules['d4rl'] = mock_d4rl

env = MockEnv()

ARGS = {
    'env': env
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
d4rl, gym, numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type dict — a dict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dict output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
load_queries_with_indices Ran
ran; our contract draft was wrong, not the code
pickxiguapi/clean-offline-rlhf/rlhf/train_reward_model.py
code served (permissive licence) · get_code("1997c70f233e5ebd")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned dict. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np

# Set random seed for reproducibility
np.random.seed(0)

# Create a small dataset
num_query = 2
len_query = 3
obs_dim = 4
action_dim = 2
total_trajectory_length = 20

dataset = {
    'observations': np.random.randn(total_trajectory_length, obs_dim).astype(np.float32),
    'actions': np.random.randn(total_trajectory_length, action_dim).astype(np.float32),
    'rewards': np.random.randn(total_trajectory_length).astype(np.float32)
}

# Create saved_indices: 2 lists of indices (for trajectory 1 and 2)
saved_indices = [
    np.array([0, 5]),      # indices for first trajectory variant
    np.array([2, 7])       # indices for second trajectory variant
]

# Create saved_labels: preference labels for each query
saved_labels = np.array([0, 1])  # 0 means prefer seq_1, 1 means prefer seq_2, -1 for unclear

ARGS = {
    'dataset': dataset,
    'num_query': num_query,
    'len_query': len_query,
    'saved_indices': saved_indices,
    'saved_labels': saved_labels,
    'label_type': 1,
    'scripted_teacher': False,
    'equivalence_threshold': 0.5,
    'modality': 'state',
    'partition_idx': None
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type dict — a dict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dict output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
normalize_states Ran
ran, honoured the contract we drafted
this paper's copy was not recorded; identical code first harvested from tinkoff-ai/CORL
pointer only · get_code("ba80f9ff81b1de5f")
How it ran

The body of this sample is withheld: this paper's copy of the code was not recorded (its link names no repository or source path), so no licence was read for it. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

reformat_video_info Ran
ran on a synthesized input
pickxiguapi/uni-rlhf-platform/uni_rlhf/datasets/dataset_utils.py
code served (permissive licence) · get_code("5b2bf6a826a95574")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned list, ndim 1, dtype family float; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import numpy as np

# Analyze the function: reformat_video_info takes video_info (a dict)
# where values are lists/arrays that can contain ints, floats, or strings.
# It iterates through indices and converts non-string values to int.

# Create sample video_info dict with lists of mixed types
video_info = {
    "video_id": ["vid_001", "vid_002"],
    "frame_count": [150.7, 240.3],
    "fps": [30, 24],
    "duration": [5.0, 10.0],
    "title": ["video_a", "video_b"]
}

ARGS = {
    "video_info": video_info
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type list, ndim 1, dtype family float — a list output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
to_dict Ran
ran on a synthesized input
pickxiguapi/uni-rlhf-platform/uni_rlhf/utils.py
code served (permissive licence) · get_code("5f825bf2f41fd48e")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned dict. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import torch.nn as nn
from sqlalchemy import Column, Integer, String, create_engine
from sqlalchemy.ext.declarative import declarative_base
from sqlalchemy.orm import sessionmaker

torch.manual_seed(0)

# Create a simple SQLAlchemy model
Base = declarative_base()

class ModelTable(Base):
    __tablename__ = 'model_table'
    
    id = Column(Integer, primary_key=True)
    name = Column(String)
    value = Column(Integer)
    description = Column(String)

# Create an instance of the model
model_instance = ModelTable(
    id=1,
    name="test_model",
    value=42,
    description="A test model instance"
)

ARGS = {
    "model_instance": model_instance
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
sqlalchemy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type dict — a dict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dict output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block

Also present in other papers' code

Of 9 samples lifted from this paper's linked repositories, 9 carry a function identity and 0 do not. Of the 9 distinct functions with an identity, 3 are also present in other papers whose code we hold (3 with at least one paper dated earlier, 0 the same month, 2 later) and 6 are only here among the 52,977 papers whose code Syntology holds (as of 2026-09-26).

Only here means: among the 52,977 papers whose code Syntology holds (51,887 with lifted samples, 1,090 through their assemblies only), as of 2026-09-26, no other paper's lifted samples or current assembly bricks contain an identical function body after normalisation.

FunctionAlso present inEarliest-datedOthers named
normalize_states 14 other papers
7 dated earlier, 7 later
2106.06860 (2021-06)
A Minimalist Approach to Offline Reinforcement Learning
2210.07105 (2022-10)
CORL: Research-oriented Deep Offline Reinforcement Learning Library
2303.06614 (2023-03)
Synthetic Experience Replay
and 11 more not named here
compute_mean_std 12 other papers
6 dated earlier, 6 later
2106.06860 (2021-06)
A Minimalist Approach to Offline Reinforcement Learning
2210.07105 (2022-10)
CORL: Research-oriented Deep Offline Reinforcement Learning Library
2303.06614 (2023-03)
Synthetic Experience Replay
and 9 more not named here
allowed_file 4 other papers
4 dated earlier
1411.4555 (2014-11)
Show and Tell: A Neural Image Caption Generator
1609.03605 (2016-09)
Detecting Text in Natural Image with Connectionist Text Proposal Network
1612.01010 (2016-12)
DeepBach: a Steerable Model for Bach Chorales Generation
and 1 more not named here

Presence, not direction: an identical function body in two papers' repositories does not say who wrote it, which came from which, or that either paper's method rests on it. Matching is exact after normalisation, so a renamed or edited function is not matched. Dates are when papers appeared, not when code was written. Only .py files the harvester lifted are covered; notebooks are not read. A new-style arXiv id YYMM.NNNN dates a paper to the month 20YY-MM; otherwise Paper.year dates it to the year; otherwise it is undated. Dates are when the paper appeared, not when the code was written. Read 2026-11-22.

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

Reinforcement Learning with Human Feedback (RLHF) has received significant attention for performing tasks without the need for costly manual reward design by aligning human preferences. It is crucial to consider diverse human feedback types and various learning methods in different environments. However, quantifying progress in RLHF with diverse feedback is challenging due to the lack of standardized annotation platforms and widely used unified benchmarks. To bridge this gap, we introduce Uni-RLHF, a comprehensive system implementation tailored for RLHF. It aims to provide a complete workflow from real human feedback, fostering progress in the development of practical problems. Uni-RLHF contains three packages: 1) a universal multi-feedback annotation platform, 2) large-scale crowdsourced feedback datasets, and 3) modular offline RLHF baseline implementations. Uni-RLHF develops a user-friendly annotation interface tailored to various feedback types, compatible with a wide range of mainstream RL environments. We then establish a systematic pipeline of crowdsourced annotations, resulting in large-scale annotated datasets comprising more than 15 million steps across 30 popular tasks. Through extensive experiments, the results in the collected datasets demonstrate competitive performance compared to those from well-designed manual rewards. We evaluate various design choices and offer insights into their strengths and potential areas of improvement. We wish to build valuable open-source platforms, datasets, and baselines to facilitate the development of more robust and reliable RLHF solutions based on realistic human feedback. The website is available at https://uni-rlhf.github.io/.

For agents

The same record, over MCP at https://syntology.ai/mcp:

get_harvested_code_for_paper("2402.02423")
get_code_for_paper("2402.02423")
have("2402.02423")

The run record, dated, one paper per request, free:

curl https://syntology.ai/api/ran/2402.02423.json

A badge for a README (the split and the date, never a ratio):

[![Syntology run record](https://syntology.ai/api/ran/2402.02423.svg)](https://syntology.ai/paper/2402.02423)

Connect an agent — have() is free.