README.md

Retroactive Reasoning Network (RRN) for Question Answering

Model Description

This model implements an Retroactive Reasoning Network (RRN) for Question Answering tasks. The RRN architecture enables multi-step reasoning through an iterative refinement process that retroactively updates hidden states.

Key Features

• Multi-step Reasoning: The model performs 3 reasoning steps to iteratively refine its predictions.
• Dynamic Reasoning Steps: Enabled - Uses a learned approach to determine the number of steps (min: 1, max: 5)
• Gating Mechanism: Selectively applies updates to hidden states.
• Delta Magnitude Constraint: Prevents destabilizing updates with a target ratio of 0.2.
• Active Memory: Stores and retrieves examples to enhance reasoning.

Usage

from transformers import AutoTokenizer
from model import EnhancedRRN_QA_Model

# Load tokenizer and model
tokenizer = AutoTokenizer.from_pretrained("will4381/rrn-qa")
model = EnhancedRRN_QA_Model("will4381/rrn-qa")

# Load custom components
import torch
import os

model.qa_head.load_state_dict(torch.load(os.path.join("will4381/rrn-qa", "qa_head.pth")))
model.retroactive_update_layer.load_state_dict(torch.load(os.path.join("will4381/rrn-qa", "retroactive_layer.pth")))
model.gating_mechanism.load_state_dict(torch.load(os.path.join("will4381/rrn-qa]", "gating_mechanism.pth")))

# If using learned dynamic steps
if os.path.exists(os.path.join("will4381/rrn-qa", "step_controller.pth")) and hasattr(model, "step_controller"):
    model.step_controller.load_state_dict(torch.load(os.path.join("will4381/rrn-qa", "step_controller.pth")))

# Example usage
inputs = tokenizer("What is the capital of France?", "Paris is the capital of France.", return_tensors="pt")
outputs = model(**inputs)

Training

This model was trained on the SQuAD dataset using a multi-step reasoning approach. The training code is included in the code directory of this repository.

To train your own model:

python code/train.py

To evaluate the model:

python code/test_model.py

Model Architecture

The RRN architecture consists of:

1. A base language model (BERT)
2. A retroactive update layer that computes delta updates
3. A gating mechanism for selective updates
4. An enhanced QA head for answer prediction
5. A step controller for dynamic reasoning steps (if enabled)

Evaluation Results

{'exact_match': 78.79848628193, 'f1': 86.94253357952118}