Tharindu Cyril Weerasooriya

HUMANLM: Simulating Users with State Alignment Beats Response Imitation

Thu, 01 Jan 2026 00:00:00 +0000

LPI-RIT at LeWiDi-2025: Improving Distributional Predictions via Metadata and Loss Reweighting with DisCo

Sat, 01 Nov 2025 00:00:00 +0000

Subasa - Adapting Language Models for Low-resourced Offensive Language Detection in Sinhala

Tue, 01 Apr 2025 00:00:00 +0000

ARTICLE: Annotator Reliability Through In-Context Learning

Sat, 01 Mar 2025 00:00:00 +0000

When humans label data for AI (like deciding if a tweet is offensive), how do we know who’s doing good work versus who’s careless or biased? Traditional methods struggle because on subjective tasks, sincere people genuinely disagree. We built ARTICLE, which uses large language models to check if an annotator’s labels are internally consistent and align with what the AI predicts. This helps identify reliable annotators without silencing minority perspectives—crucial for building fair AI systems that don’t just reflect the majority view.

Blind Spot Navigation in Large Language Model Reasoning with Thought Space Explorer

Wed, 01 Jan 2025 00:00:00 +0000

Note

Presenting at NeurIPS (Math-AI Workshop)
Date: December 6, 2025
Time: Sat 3:30 p.m. - 4:15 p.m. Location: NeurIPS 2025 - Workshop Upper Level Ballroom 6A Session Type: Poster Session

MultiGA: Leveraging Multi-Source Seeding in Genetic Algorithms

Wed, 01 Jan 2025 00:00:00 +0000

Different LLMs (like GPT-4, Claude, DeepSeek) are good at different things. Instead of picking one model and hoping for the best, MultiGA uses genetic algorithms—inspired by natural evolution—to combine outputs from multiple LLMs. It starts by generating diverse answers from several models, evaluates which ones are best, then “breeds” improved solutions by mixing and mutating them over multiple generations. We tested this on text-to-SQL generation, trip planning, and grad-level science questions, and found it matches or beats the best single model. Think of it as getting a second (and third, and fourth) opinion from different AI experts, then synthesizing their best ideas.

ProRefine: Inference-Time Prompt Refinement with Textual Feedback

Wed, 01 Jan 2025 00:00:00 +0000

When multiple AI agents work together (like one agent planning and another executing), they rely heavily on good prompts. But writing perfect prompts is hard, and even small mistakes can cascade through the system, causing failures. ProRefine solves this by creating a feedback loop: one AI agent evaluates how well the prompt worked and suggests improvements, then another agent refines the prompt based on that feedback—all happening automatically at inference time without any training.

We tested ProRefine on mathematical reasoning tasks and it improved accuracy by 3-37% compared to standard methods. Even better, it lets smaller, cheaper AI models perform nearly as well as much larger, expensive ones. This makes powerful AI more accessible and cost-effective, especially for complex multi-agent systems that are becoming critical in commercial applications.

Rater Cohesion and Quality from a Vicarious Perspective

Fri, 01 Nov 2024 00:00:00 +0000

When labeling controversial content (like tweets about politics), people’s own beliefs strongly influence what they find offensive. We used “vicarious annotation”—asking people not just “Is this offensive to you?” but also “Would a Democrat/Republican/Independent find this offensive?” This revealed fascinating patterns: Republicans were the worst at predicting how others would react, and disagreement spiked on hot-button issues like gun control and abortion. Understanding these patterns helps us build content moderation AI that doesn’t just reflect one group’s values while silencing others.

Harnessing Business and Media Insights with Large Language Models

Mon, 01 Jan 2024 00:00:00 +0000

Vicarious Offense and Noise Audit of Offensive Speech Classifiers: Unifying Human and Machine Disagreement on What is Offensive

Sat, 02 Dec 2023 00:00:00 +0000

We ran a massive experiment: 9 different AI content moderation systems analyzed 92 million YouTube comments about US politics. The results were shocking—different AI systems flagged wildly different content as offensive, with almost no consistency. When we asked humans to label the same content, political identity was a huge factor: Democrats and Republicans disagreed strongly on what’s offensive, especially on hot topics like abortion and gun rights. This proves that “offensiveness” isn’t a fact that AI can learn—it’s a subjective judgment shaped by values. Current content moderation practices that treat one group’s perspective as “truth” are fundamentally unfair.

Example from the Annotation

Disagreement Matters: Preserving Label Diversity by Jointly Modeling Item and Annotator Label Distributions with DisCo

Sat, 01 Jul 2023 00:00:00 +0000

Subjective Crowd Disagreements for Subjective Data: Uncovering Meaningful CrowdOpinion with Population-level Learning

Sat, 01 Jul 2023 00:00:00 +0000

Most AI training forces annotators to agree on a single “correct” label, erasing legitimate differences of opinion. But when humans disagree (like on whether a tweet is offensive), that disagreement is often meaningful—not noise. CrowdOpinion solves this by clustering similar content and predicting the full distribution of opinions people might have, rather than picking one “winner.” We tested it on Twitter, Reddit, Gab, and Facebook data, where disagreement is common. This approach ensures minority viewpoints aren’t silenced, which is crucial for building fair AI that respects diverse perspectives—especially important for content moderation and other subjective decisions.

Findings from the Bambara - French Machine Translation Competition (BFMT 2023)

Mon, 01 May 2023 00:00:00 +0000

Vicarious Offense and Noise Audit of Offensive Speech Classifiers

Wed, 01 Feb 2023 00:00:00 +0000

Annotator Response Distributions as a Sampling Frame

Sat, 01 Jan 2022 00:00:00 +0000

Improving Label Quality by Joint Probabilistic Modeling of Items and Annotators

Sat, 01 Jan 2022 00:00:00 +0000

Improving Label Quality by Jointly Modeling Items and Annotators

Fri, 01 Jan 2021 00:00:00 +0000

Neighborhood-based pooling for population-level label distribution learning

Wed, 01 Jan 2020 00:00:00 +0000

Cloud-Based Open Source Primary Care Electronic Patient Record System for Sri Lankan Citizens

Tue, 01 Jan 2019 00:00:00 +0000