reSAID Lab

Coding Agents and Operational Safety

Mon, 16 Nov 2026 00:00:00 +0000

Overview

Autonomous coding agents built on large language models are wired directly into development workflows: they edit files, run commands, configure environments, and fix bugs with growing autonomy. Most safety evaluations of these tools focus on explicitly malicious prompts, but we argue this misses the larger and more common danger: agents that fail during ordinary, goal-directed work through destructive operations, constraint violations, authorization bypasses, and silent errors that surface only after damage is done.

What Breaks When LLMs Code? Characterizing Operational Safety Failures of Agentic Code Assistants

Mon, 16 Nov 2026 00:00:00 +0000

ReShift: Aha-Moment-Driven Reasoning-Level Backdoor Attacks on Vision–Language Models

Tue, 08 Sep 2026 00:00:00 +0000

Trustworthy LLMs and VLMs

Tue, 08 Sep 2026 00:00:00 +0000

Overview

Large language and vision-language models are deployed in settings where biased, inconsistent, or manipulated behavior can affect users, yet their internals are often unavailable or hard to inspect. We develop methods that expose and characterize such hidden failures, treating trustworthiness as a property that must be tested for rather than assumed — and connecting each testing method to a concrete path for mitigation or defense.

A recurring theme in our work is that trustworthiness must account for a model’s reasoning process, not only its final answer. Attacks and guardrails that operate on outputs alone tend to leave reasoning traces that are inconsistent or easy to flag, but as models increasingly expose their chain-of-thought, the reasoning itself becomes both a new attack surface and a new opportunity for defense. We study how bias and backdoor threats propagate through model behavior, how to characterize them with principled signals, and how to build safeguards that hold up against adaptive adversaries.

Alif Al Hasan

Sat, 25 Jul 2026 00:00:00 +0000

About

Alif Al Hasan is a Ph.D. student in the Department of Computer and Data Sciences at Case Western Reserve University, working under the supervision of Prof. Sumon Biswas at the reSAID Lab. He earned his Bachelor’s and Master’s degrees in Computer Science and Engineering from Jahangirnagar University. His research operates at the intersection of Software Engineering and AI, focusing on the operational safety of autonomous LLM agents.

Contact

Email: alifal.hasan@case.edu
Website: alifalhasan.github.io
GitHub: alifalhasan

LLM Reasoning and Planning

Mon, 13 Jul 2026 00:00:00 +0000

Overview

Large language models can appear to reason, yet generation is autoregressive: each token is chosen from the immediate context, one step at a time. This local view is powerful, but it explains familiar failure modes, such as reasoning that drifts, contradicts itself, takes redundant detours, or commits early to a path that later proves wrong. We study how to make model reasoning globally coherent, efficient, and trustworthy by helping a model decide where it is going before it takes the next step.

Plan Then Action: High-Level Planning Guidance Reinforcement Learning for LLM Reasoning

Mon, 13 Jul 2026 00:00:00 +0000

ReShift backdoor paper accepted to ECCV 2026

Thu, 18 Jun 2026 00:00:00 +0000

Our paper proposing ReShift, a reasoning-level backdoor framework for Vision–Language Models, was accepted to ECCV 2026 in Malmö, Sweden.

Co-Chair of the AAAI Fall Symposium on Trustworthy Agentic Systems (TAS 2026)

Mon, 15 Jun 2026 00:00:00 +0000

Sumon Biswas is serving as Co-Chair of the AAAI Fall Symposium on Trustworthy Agentic Systems (TAS 2026), November 5-7, 2026, in Arlington, Virginia.

What Actually Breaks When LLMs Write Code?

Thu, 11 Jun 2026 00:00:00 +0000

Most conversations about the safety of coding agents revolve around adversarial scenarios: prompt injection, jailbreaks, malicious instructions hidden in a README. Those threats are real. But after watching these tools work — and occasionally watching them wreck a working environment while “fixing” a unit test — we kept returning to a more uncomfortable question: what goes wrong when nobody is attacking, and the agent is simply trying to help?

Our new preprint, What Breaks When LLMs Code?, led by our Ph.D. student Alif Al Hasan, is an attempt to answer that question with evidence rather than anecdotes. We call this operational safety: the safety of an agent during benign, goal-directed, everyday use.

Teaching LLMs to Plan Before They Act

Wed, 10 Jun 2026 00:00:00 +0000

If you have ever watched a language model reason its way through a hard math problem, you have probably seen it wander. The chain of thought starts off promising, circles back on itself, re-derives something it already knew, and occasionally talks itself out of a correct intermediate result. The final answer may still be right, but the path there is long, redundant, and hard to trust.

Our ICML 2026 paper, Plan Then Action, starts from a simple diagnosis of why this happens: autoregressive generation is local. At every step the model decides only what token comes next, so the reasoning process is essentially a sequence of small, greedy decisions. There is no global plan — nothing that commits the model to a strategy before it starts executing one. Tree search and reinforcement learning can partially compensate, but they are expensive and still operate over the same token-level process.

Agentic code safety paper accepted to ASE 2026

Fri, 29 May 2026 00:00:00 +0000

Our empirical study characterizing operational safety failures of LLM-based coding agents was accepted to ASE 2026 in Munich, Germany.

Copilot teaching tutorial at the NSF AI Unlocked workshop

Fri, 15 May 2026 00:00:00 +0000

Sumon Biswas presented the hands-on tutorial ‘Teaching Code-Generation Courses with GitHub Copilot’ at the NSF AI Unlocked workshop, hosted by CU Boulder Research Computing with ACCESS and the NAIRR Pilot.

Ruksaar Shaik defends M.S. project and graduates

Fri, 15 May 2026 00:00:00 +0000

Ruksaar Shaik successfully defended her M.S. project and graduated in May 2026. Her work focused on building domain-specific LLM models for detecting intimate partner violence (IPV) from natural-language social media posts.

Ruksaar Shaik presents IPV-detection poster at the CTSC AI Summit

Sun, 10 May 2026 00:00:00 +0000

Ruksaar Shaik presented her poster on detecting intimate partner violence (IPV) with large language models at the CTSC AI Summit in May 2026.

Plan Then Action accepted to ICML 2026

Wed, 01 Apr 2026 00:00:00 +0000

Our paper on high-level planning guidance reinforcement learning for LLM reasoning was accepted to ICML 2026 in Seoul, South Korea.

General Chair of the LLMTrust workshop

Sun, 01 Feb 2026 00:00:00 +0000

Sumon Biswas is serving as General Chair of the International Workshop on Trustworthy Large Language Models for Software Engineering (LLMTrust).

Bias Testing and Mitigation in Black Box LLMs using Metamorphic Relations

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

UCITE Learning Fellow

Mon, 01 Dec 2025 00:00:00 +0000

Sumon Biswas was accepted as a UCITE Learning Fellow at Case Western Reserve University.

CTSC pilot grant for AI-assisted dating-violence prevention

Mon, 01 Sep 2025 00:00:00 +0000

Sumon Biswas was awarded a CWRU CTSC pilot grant to study fairness assessment and improvement for AI-enabled detection of dating violence in youth digital communication, in collaboration with University Hospitals.

Avyukth Sai Rangarajan

Fri, 15 Aug 2025 00:00:00 +0000

Devak Pardasani

Fri, 15 Aug 2025 00:00:00 +0000

Phat Dang

Fri, 15 Aug 2025 00:00:00 +0000

Ram Aryan Mallampati

Fri, 15 Aug 2025 00:00:00 +0000

Sam Lin

Fri, 15 Aug 2025 00:00:00 +0000

Zelan Eroz Espanto

Fri, 15 Aug 2025 00:00:00 +0000

Invited talk at Kent State AUTOBOT

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

Sumon Biswas gave an invited talk, Engineering Responsible AI: From Fairness to Long-term Impact, at the Robotics and Autonomous Systems (AUTOBOT) Program at Kent State University.

FairSense: Long-Term Fairness Analysis of ML-Enabled Systems

Thu, 01 May 2025 00:00:00 +0000

Long-Term Fairness and ML Safety

Thu, 01 May 2025 00:00:00 +0000

Overview

Many ML-enabled systems operate in dynamic environments: the system’s decisions change the environment, and those changes feed back into its future inputs. Certain self-reinforcing loops can amplify errors, entrench bias, and cause fairness violations in the long term even when immediate outcomes are fair. In predictive policing, for example, a model that flags a neighborhood as high-crime sends more patrols there, producing more recorded arrests, which the model reads as even higher crime. The same pattern appears in loan approvals that affect credit scores and in medical risk scoring that influences treatment access.

New Faculty Symposium at ICSE 2025

Thu, 01 May 2025 00:00:00 +0000

Sumon Biswas attended the New Faculty Symposium at ICSE 2025 in Ottawa, Canada.

Data science pipelines work featured on New Books Network

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

The story behind The Art and Practice of Data Science Pipelines was featured on the New Books Network podcast.

FairSense accepted to ICSE 2025

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

Our paper on long-term fairness analysis of ML-enabled systems was accepted to ICSE 2025 in Ottawa, Canada.

Ali Nawaf