Amirreza Vishteh

I am a researcher focused on AI Safety, Trustworthy AI, and Large Language Models (LLMs). I graduated with a Bachelor's degree from Iran University of Science and Technology (IUST) and am currently pursuing a Master's degree in Computer Engineering at Sharif University of Technology.

Psychological Health Chatbot: Enhancing Mental Well-being with AI

Model

Overview

The Psychological Health Chatbot is designed to assist individuals in their mental health journey by leveraging advanced natural language processing (NLP) techniques. Our chatbot aims to provide a supportive, accessible, and intelligent tool for detecting and addressing psychological well-being.

📄 Read our full paper here: ACL Anthology

Team Members

Sadegh Jafari
Mohammad Erfan Zare
Amirreza Vishteh
Mirzae Melike
Zahra Amiri
Sima Mohammadparast
Sauleh Eetemadi

Data Collection
Model Training
Evaluation
Results
Conclusion
Future Work

1. Data Collection

To ensure the chatbot effectively understands and responds to mental health queries, we compiled a diverse dataset of conversations related to psychological well-being. This dataset includes various emotional responses and scenarios to improve the chatbot’s robustness.

Data Collection

2. Model Training

We fine-tuned a pre-trained transformer-based model to enhance contextual understanding and response accuracy. Key aspects of training included:

Optimizing hyperparameters
Incorporating advanced emotion recognition techniques
Improving contextual relevance in responses

Model Training

3. Evaluation

Our chatbot underwent rigorous evaluation through user testing. We assessed:

Emotion detection accuracy
User satisfaction rates
Response relevance and coherence

Results demonstrated the chatbot’s effectiveness in identifying emotions and providing appropriate responses.

4. Results

Our final evaluation yielded promising results, with user satisfaction rates reaching [insert percentage]%. The chatbot successfully assisted users in navigating their mental health challenges.

Results Results

Conclusion

The Psychological Health Chatbot represents a significant step in integrating AI-driven solutions for mental health support. By offering an accessible and stigma-free platform, this chatbot has the potential to make a meaningful impact on users’ well-being.

Future Work

Future developments will focus on:

Enhancing emotional detection capabilities
Integrating personalized feedback mechanisms
Improving multi-turn conversational depth

📄 For more details, read our paper: ACL Anthology

2026 1
2025 11
2024 4
2023 2
2021 2
2020 3

2026

MECP-GAP: Mobility-Aware MEC Planning via GNNs

5 minute read

2025

A Watchdog Model for Backdoored LoRA Adapters

1 minute read

Overview Most simple backdoor checks look for a specific trigger phrase. That breaks the moment the trigger is unknown or paraphrased. This project asks a di...

The Sentinel’s Dilemma: Guarding AI from Hidden Threats

2 minute read

The Sentinel’s Dilemma: Who Guards Our AI Guardians?

Turning a Backdoor’s Step Function Into a Ramp

1 minute read

Overview A backdoor behaves like a step function: invisible until its trigger fires, fully present the instant it does. That makes it a bad target for anythi...

Auditing Backdoors: From a Flag to a Causal Proof

1 minute read

Overview A model that gets flagged as “probably backdoored” is not the same as a model whose exact trigger and payload have been proven, causally, to be conn...

Re-running a Vulnerability Detection Benchmark, More Carefully

1 minute read

Overview A widely cited graph neural network for detecting vulnerable code is reproduced from scratch on the authors’ own released data, then put under three...

Toward a Field-Based View of LLM Backdoors

1 minute read

Overview Existing LLM backdoor scanners mostly search for an object: a trigger string, a target string, or a memorized leaked example. This project explores ...

Building a Weakness Zoo to Stress-Test a Backdoor Scanner

1 minute read

Overview A backdoor scanner that only gets tested against easy, textbook backdoors will look better than it is. This project builds a “weakness zoo” — a deli...

Starting Point: Vision-Language Models for Dental X-rays

less than 1 minute read

Overview Dental panoramic X-rays are read manually by clinicians, and there is growing interest in whether general-purpose medical vision-language models can...

Comparing Five Ways to Catch a Backdoor, Live

1 minute read

Overview Most backdoor-detection papers evaluate on large models behind a GPU cluster, which makes it hard to build intuition about when a detection signal a...

Understanding BackdoorBench: A Comprehensive Benchmark for AI Security

2 minute read

Unveiling BackdoorBench: A Critical Benchmark for AI Security

FTT-NAS: Implementation and Review

3 minute read

2024

Wav2Vec2 Sentiment Analysis Using Shemo Dataset

1 minute read

Overview In this project, we fine-tuned the Wav2Vec2 model to perform sentiment analysis based on both voice features and text transcripts from the Shemo da...