UniRRM-14B API & Inference Endpoint

Overview

UniRRM-14B is a unified reasoning reward model that supports multiple languages (103 languages) and multiple evaluation paradigms (pairwise, listwise, and pointwise) in a single model. It is built on Qwen3-14B and trained with a two-stage pipeline (SFT + GRPO) on the MixReward dataset.

This model is introduced in the following paper, accepted at ICML 2026 (the 43rd International Conference on Machine Learning):

UniRRM: Unified Reasoning Reward Models Across Languages and Evaluation Paradigms [Paper]

Key Features

🌍 103 Languages: Trained on multilingual data spanning 103 languages across 6 domains
🔀 Unified Evaluation Paradigms: Supports pairwise, listwise, and pointwise evaluation in a single model
🧠 Adaptive Rubric Generation: Dynamically generates task-generic and instruction-specific evaluation criteria through a staged reasoning chain
⚡ Structured Reasoning: Follows a three-stage reasoning pipeline — Deep Analysis → Adaptive Rubric Generation → Detailed Evaluation

Reasoning Workflow

UniRRM follows a structured three-stage reasoning chain:

Deep Analysis (𝒛): Identifies task intent, potential risks, core evaluation objectives, and strict constraints
Adaptive Rubric Generation (𝒓): Produces both task-generic criteria (broadly applicable) and instruction-specific criteria (tailored to user query), each on a 1–5 scoring scale
Detailed Evaluation (𝒆): Applies generated rubrics to judge candidate responses with per-criterion scoring and final judgment

Quick Start with vLLM

The following example demonstrates pairwise evaluation using vLLM offline inference. To switch to other evaluation paradigms, simply adjust the number of <Response> blocks in the user prompt:

Pairwise: 2 responses (<Response1>, <Response2>)
Listwise: 4 responses (<Response1> through <Response4>)
Pointwise: 1 response (<Response1>), optionally with a <Reference_Answer> block

python
import json
import re
from vllm import LLM, SamplingParams
from transformers import AutoTokenizer

MODEL_NAME = "SUSTech-NLP/UniRRM-14B"

# ---------- 1. Load model ----------
tokenizer = AutoTokenizer.from_pretrained(MODEL_NAME)
llm = LLM(model=MODEL_NAME, tensor_parallel_size=2, max_model_len=16384)
sampling_params = SamplingParams(temperature=0, max_tokens=4096, repetition_penalty=1.05)

# ---------- 2. Build prompt ----------
SYSTEM_PROMPT = """
You are a multilingual evaluation expert, responsible for conducting rigorous, objective, and multi-dimensional evaluations of responses generated for User Input. Your evaluation must strictly follow the step-by-step process outlined below:

### Phase 1: Deep Analysis
Before evaluating, perform a comprehensive analysis of the User Input to establish a robust baseline:
1. **Identify potential risks**: Analyze the User Input to identify any potential safety, legal, offensive, or ethical risks.
2. **Identify task type**: Identify the primary task type (e.g., chat, reasoning, code generation, translation, or creative writing).
3. **Analyze core requirements (task-dependent)**: Define the fundamental evaluation dimensions that any correct response must satisfy.
4. **Analyze specific requirements**: Identify additional constraints or expectations unique to the User Input.
5. **Predict response content**: Summarize the expected content or core objectives of a correct response.

### Phase 2: Dynamic Rubric Generation
1. Generate a set of evaluation rubrics tailored to the user inputs and responses, with a 1-5 scoring criterion for each rubric.
2. If any safety, legal, or ethical risks are detected, include a Safety rubric as the highest-priority dimension.
3. Ensure rubrics comprehensively cover all critical aspects of the response.

### Phase 3: Detailed Evaluation
For each rubric, evaluate the response:
1. **Evidence Extraction**: Identify specific passages that meet or fail to meet the rubric requirements.
2. **Gap Analysis**: Determine why the response did not achieve a perfect score (5).
3. **Scoring**: Assign a score from 1 to 5.

### OUTPUT FORMAT
{
  "Analysis_process": "Concise summary of the analysis.",
  "rubrics": [{"name": "String", "description": "Rubric definition"}],
  "evaluations": [{"response_id": "String", "explanation": "Summary", "final_score": "Float"}],
  "best_id": "ID of the winner"
}
""".strip()

question = "Explain the concept of recursion in programming."
response_a = "Recursion is when a function calls itself to solve smaller subproblems. A base case stops the recursion, and each recursive call works on a reduced version of the original problem. For example, calculating factorial: factorial(n) = n * factorial(n-1), with factorial(0) = 1 as the base case."
response_b = "Recursion means repeating something. In programming, it is used sometimes."

user_prompt = f"""
<User_Input>
{question}
</User_Input>

<Response1>
{response_a}
</Response1>

<Response2>
{response_b}
</Response2>
"""

messages = [
    {"role": "system", "content": SYSTEM_PROMPT},
    {"role": "user", "content": user_prompt},
]
prompt = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)

# ---------- 3. Generate ----------
outputs = llm.generate([prompt], sampling_params)
raw_output = outputs[0].outputs[0].text
print(raw_output)

# ---------- 4. Parse output ----------
def parse_unirm_output(raw_output: str) -> dict:
    """Parse UniRRM's JSON output to extract scores and best_id."""
    text = raw_output
    # Strip " in text:
        text = text.split("</think>")[-1].strip()

    # Extract JSON from markdown code block or raw text
    code_block = re.search(r"```(?:json)?\s*(\{.*?\})\s*```", text, re.DOTALL)
    if code_block:
        json_str = code_block.group(1)
    else:
        start, end = text.find("{"), text.rfind("}")
        if start != -1 and end != -1:
            json_str = text[start : end + 1]
        else:
            return {"error": "No JSON found in output"}

    try:
        return json.loads(json_str)
    except json.JSONDecodeError:
        match = re.search(r'"final_score"\s*:\s*"?(\d+(?:\.\d+)?)"?', json_str)
        if match:
            return {"final_score": float(match.group(1))}
        return {"error": "Failed to parse JSON"}

result = parse_unirm_output(raw_output)
print(f"Best response: {result.get('best_id')}")
for evaluation in result.get("evaluations", []):
    print(f"  {evaluation['response_id']}: score={evaluation['final_score']}")

Training

UniRRM-14B is trained using a two-stage pipeline:

Stage 1: Supervised Fine-Tuning (SFT)

Base model: Qwen3-14B
Training data: UniRRM-SFT (35,749 samples distilled from GPT-OSS-120B)
Objective: Initialize structured reasoning capabilities (analysis → rubric generation → evaluation)

Stage 2: Reinforcement Learning with GRPO

Training data: UniRRM-RL (32,832 samples)
Algorithm: Group Relative Policy Optimization (GRPO)
Composite reward: R = 0.8 × r_fmt + 0.15 × r_acc + 0.05 × r_rubric
- Format Reward (r_fmt): Ensures structured output compliance
- Outcome Consistency Reward (r_acc): Binary reward for correct final judgment
- Rubric Quality Reward (r_rubric): Teacher model (Qwen3-Max) evaluates rubric quality (1–5)
Hyperparameters: lr=1e-6, weight_decay=0.01, batch_size=1024, epochs=2, kl_coef=0.001, rollout=5
Hardware: 16 × NVIDIA H100 80GB GPUs

Model Details

Table with columns: Attribute, Value
Attribute	Value
Architecture	Qwen3ForCausalLM
Parameters	~14B
Precision	bfloat16
Hidden Size	5120
Num Layers	40
Num Attention Heads	40
Num KV Heads	8 (GQA)
Max Position Embeddings	40960

📄 Paper: UniRRM: Unified Reasoning Reward Models Across Languages and Evaluation Paradigms (ICML 2026)
🤖 UniRRM-8B: SUSTech-NLP/UniRRM-8B
📊 MixReward Dataset: SUSTech-NLP/MixReward (64,528 samples, 103 languages)
📊 UniRRM-SFT Dataset: SUSTech-NLP/UniRRM-SFT (35,749 SFT samples)
📊 UniRRM-RL Dataset: SUSTech-NLP/UniRRM-RL (32,832 RL samples)

Citation

bibtex
@inproceedings{
lai2026unirrm,
title={Uni{RRM}: Unified Reasoning Reward Models Across Languages and Evaluation Paradigms},
author={Peng Lai and Yichao Du and Junchao Wu and Weibo Gao and Linan Yue and Longyue Wang and Weihua Luo and Derek F. Wong and Guanhua Chen},
booktitle={Forty-third International Conference on Machine Learning},
year={2026},
url={https://openreview.net/forum?id=laiK6TlhL2}
}

License

This model is released under the Apache 2.0 License.

python

import json
import re
from vllm import LLM, SamplingParams
from transformers import AutoTokenizer

MODEL_NAME = "SUSTech-NLP/UniRRM-14B"

# ---------- 1. Load model ----------
tokenizer = AutoTokenizer.from_pretrained(MODEL_NAME)
llm = LLM(model=MODEL_NAME, tensor_parallel_size=2, max_model_len=16384)
sampling_params = SamplingParams(temperature=0, max_tokens=4096, repetition_penalty=1.05)

# ---------- 2. Build prompt ----------
SYSTEM_PROMPT = """
You are a multilingual evaluation expert, responsible for conducting rigorous, objective, and multi-dimensional evaluations of responses generated for User Input. Your evaluation must strictly follow the step-by-step process outlined below:

### Phase 1: Deep Analysis
Before evaluating, perform a comprehensive analysis of the User Input to establish a robust baseline:
1. **Identify potential risks**: Analyze the User Input to identify any potential safety, legal, offensive, or ethical risks.
2. **Identify task type**: Identify the primary task type (e.g., chat, reasoning, code generation, translation, or creative writing).
3. **Analyze core requirements (task-dependent)**: Define the fundamental evaluation dimensions that any correct response must satisfy.
4. **Analyze specific requirements**: Identify additional constraints or expectations unique to the User Input.
5. **Predict response content**: Summarize the expected content or core objectives of a correct response.

### Phase 2: Dynamic Rubric Generation
1. Generate a set of evaluation rubrics tailored to the user inputs and responses, with a 1-5 scoring criterion for each rubric.
2. If any safety, legal, or ethical risks are detected, include a Safety rubric as the highest-priority dimension.
3. Ensure rubrics comprehensively cover all critical aspects of the response.

### Phase 3: Detailed Evaluation
For each rubric, evaluate the response:
1. **Evidence Extraction**: Identify specific passages that meet or fail to meet the rubric requirements.
2. **Gap Analysis**: Determine why the response did not achieve a perfect score (5).
3. **Scoring**: Assign a score from 1 to 5.

### OUTPUT FORMAT
{
  "Analysis_process": "Concise summary of the analysis.",
  "rubrics": [{"name": "String", "description": "Rubric definition"}],
  "evaluations": [{"response_id": "String", "explanation": "Summary", "final_score": "Float"}],
  "best_id": "ID of the winner"
}
""".strip()

question = "Explain the concept of recursion in programming."
response_a = "Recursion is when a function calls itself to solve smaller subproblems. A base case stops the recursion, and each recursive call works on a reduced version of the original problem. For example, calculating factorial: factorial(n) = n * factorial(n-1), with factorial(0) = 1 as the base case."
response_b = "Recursion means repeating something. In programming, it is used sometimes."

user_prompt = f"""
<User_Input>
{question}
</User_Input>

<Response1>
{response_a}
</Response1>

<Response2>
{response_b}
</Response2>
"""

messages = [
    {"role": "system", "content": SYSTEM_PROMPT},
    {"role": "user", "content": user_prompt},
]
prompt = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)

# ---------- 3. Generate ----------
outputs = llm.generate([prompt], sampling_params)
raw_output = outputs[0].outputs[0].text
print(raw_output)

# ---------- 4. Parse output ----------
def parse_unirm_output(raw_output: str) -> dict:
    """Parse UniRRM's JSON output to extract scores and best_id."""
    text = raw_output
    # Strip " in text:
        text = text.split("</think>")[-1].strip()

    # Extract JSON from markdown code block or raw text
    code_block = re.search(r"```(?:json)?\s*(\{.*?\})\s*```", text, re.DOTALL)
    if code_block:
        json_str = code_block.group(1)
    else:
        start, end = text.find("{"), text.rfind("}")
        if start != -1 and end != -1:
            json_str = text[start : end + 1]
        else:
            return {"error": "No JSON found in output"}

    try:
        return json.loads(json_str)
    except json.JSONDecodeError:
        match = re.search(r'"final_score"\s*:\s*"?(\d+(?:\.\d+)?)"?', json_str)
        if match:
            return {"final_score": float(match.group(1))}
        return {"error": "Failed to parse JSON"}

result = parse_unirm_output(raw_output)
print(f"Best response: {result.get('best_id')}")
for evaluation in result.get("evaluations", []):
    print(f"  {evaluation['response_id']}: score={evaluation['final_score']}")

Attribute

Value

Architecture

Qwen3ForCausalLM

Parameters

~14B

Precision

bfloat16

Hidden Size

5120

Num Layers

Num Attention Heads

Num KV Heads

8 (GQA)

Max Position Embeddings

40960

bibtex

@inproceedings{
lai2026unirrm,
title={Uni{RRM}: Unified Reasoning Reward Models Across Languages and Evaluation Paradigms},
author={Peng Lai and Yichao Du and Junchao Wu and Weibo Gao and Linan Yue and Longyue Wang and Weihua Luo and Derek F. Wong and Guanhua Chen},
booktitle={Forty-third International Conference on Machine Learning},
year={2026},
url={https://openreview.net/forum?id=laiK6TlhL2}
}

UniRRM-14B

README

Overview

Key Features

Reasoning Workflow

Quick Start with vLLM

Training

Stage 1: Supervised Fine-Tuning (SFT)

Stage 2: Reinforcement Learning with GRPO

Model Details

Citation

License

Explore FriendliAI today

README

Overview

Key Features

Reasoning Workflow

Quick Start with vLLM

Training

Stage 1: Supervised Fine-Tuning (SFT)

Stage 2: Reinforcement Learning with GRPO

Model Details

Citation

License

UniRRM-14B

README

Overview

Key Features

Reasoning Workflow

Quick Start with vLLM

Training

Stage 1: Supervised Fine-Tuning (SFT)

Stage 2: Reinforcement Learning with GRPO

Model Details

Related Resources

Citation

License

Explore FriendliAI today

README

Overview

Key Features

Reasoning Workflow

Quick Start with vLLM

Training

Stage 1: Supervised Fine-Tuning (SFT)

Stage 2: Reinforcement Learning with GRPO

Model Details

Related Resources

Citation

License