Powered by DeepSeek-R1 & Gemini

Consensus Analysis in
Academic Peer Review

Using Search-Augmented Large Language Models to automatically extract critiques, detect disagreements, and generate comprehensive meta-reviews.

Critique Extraction
Gemini 2.5 Flash
Evidence Retrieval
Multi-Source Search
Reasoning & Resolution
DeepSeek-R1

Powerful Features for Peer Review Analysis

Our comprehensive pipeline handles every aspect of peer review consensus analysis with state-of-the-art AI models.

Critique Extraction

Automatically extract and categorize critique points from peer reviews into Methodology, Experiments, Clarity, Significance, and Novelty.

Disagreement Detection

Identify conflicts and disagreements between reviewers with quantified disagreement scores from 0 to 1.

Search-Augmented Verification

Retrieve supporting or contradicting evidence from Semantic Scholar, arXiv, Google Scholar, and Tavily.

Evidence-Based Resolution

Resolve reviewer disagreements using AI reasoning powered by DeepSeek-R1 with retrieved evidence.

Meta-Review Generation

Generate comprehensive meta-reviews that synthesize all analyses with actionable recommendations.

Fast & Efficient

Process multiple reviews simultaneously with optimized API calls and intelligent rate limiting.

Secure API with rate limiting
Real-time progress tracking

The Analysis Pipeline

A systematic five-stage workflow that transforms peer reviews into actionable consensus insights.

01
Gemini 2.5 Flash Lite

Critique Extraction

Extract structured critique points from peer reviews and categorize them into five key aspects.

MethodologyExperimentsClaritySignificanceNovelty
02
Gemini 2.5 Flash Lite

Disagreement Detection

Compare all review pairs to identify conflicts with quantified disagreement scores.

Output: Disagreement Score: 0.0 - 1.0
03
LangChain Multi-Search

Search & Retrieval

Search academic sources to find evidence supporting or contradicting the critiques.

Semantic ScholararXivGoogle ScholarTavily
04
DeepSeek-R1

Disagreement Resolution

Use advanced reasoning to validate critiques and resolve disagreements based on evidence.

Output: Accepted & Rejected Critique Points
05
DeepSeek-R1

Meta-Review Generation

Generate a comprehensive meta-review synthesizing all analyses with recommendations.

Output: Structured Meta-Review Document

How It Works

Get started with MetaSearch in three simple steps.

1

Input Your Data

Provide the paper title, abstract, and peer reviews in JSON format. You can use our example or enter your own data.

2

AI Processing

Our pipeline processes the reviews through 5 stages: extraction, detection, retrieval, resolution, and generation.

3

Get Results

Receive structured critique points, disagreement analysis, retrieved evidence, and a comprehensive meta-review.

output.json
{
  "request_id": "req_1709847362.5",
  "paper_title": "Learning Disentangled Representations...",
  "critique_points": [
    {
      "Methodology": ["Promising approach to disentanglement..."],
      "Experiments": ["Needs comparison with more baselines..."],
      "Clarity": ["Well-written with clear motivation..."],
      "Significance": ["Addresses important problem..."],
      "Novelty": ["Novel use of auxiliary variables..."]
    }
  ],
  "disagreements": [
    {
      "review_pair": [0, 1],
      "disagreement_score": 0.45,
      "disagreement_details": {...}
    }
  ],
  "meta_review": "## Meta-Review\n\n### Summary..."
}

Try It Now

Enter a paper's details and reviews to see the full consensus analysis pipeline in action.

Input

Results

Results will appear here after running the pipeline