Razvan-Gabriel Dumitru

NeuroTS-Net: Multi-Class Semantic Segmentation of Pediatric Brain Tumors in Multi-Modal MRI

Darius Petelează, Razvan-Gabriel Dumitru, Bogdan Neamtu, Árpád Gellért et autres

Pediatric brain tumors are a leading cause of cancer-related mortality in children, and their small, rare, and often low-contrast subregions make accurate manual delineation challenging. Reliable automated segmentation is therefore needed to support diagnosis, treatment planning, and response assessment. Accordingly, we introduce …

Root-Cause Attribution Is a Search Problem: Continual Search for Long-Horizon Agent Failures

Harsh Raj, David Lee, Anas Mahmoud, Renxiong Wang et autres

The increasing deployment of AI agents in long-horizon tasks yields massive execution logs. Diagnosing failures within these records is crucial for reliability, as it transforms outcome-level signals into actionable interventions. The sheer scale of the data renders human review impractical, driving the …

Model or Harness? An Interaction-Centric Taxonomy for Localizing Agent Failures

Harsh Raj, Vipul Gupta, Anas Mahmoud, Razvan-Gabriel Dumitru et autres

Existing evaluations often reduce agent failures to system-level outcomes, obscuring where the fault originated and which intervention would improve the agent system. This creates a repair-assignment problem: the same visible failure may call for model post-training, harness engineering, environment redesign, or benchmark …

CRAFT: Clustering Rubrics to Diagnose Weak LLM Capabilities and Generate Targeted Fine-Tuning Data

Gupta Vk, Zihao Wang, Razvan-Gabriel Dumitru, MohammadHossein Rezaei et autres

Evaluations should do more than measure a models current performance. They should tell us what to fix for the next model iteration and provide a way to generate targeted post training data. Most evaluation pipelines identify weak examples, topics, or categories, but …

Rubric-Guided Self-Distillation: Post-Training Without Rubric Verifiers

MohammadHossein Rezaei, Anas Mahmoud, Zihao Wang, Utkarsh Tyagi et autres

Rubrics have emerged as an alternative to RLVR in open-ended domains where a single ground-truth final answer is not available. Existing rubric-based training methods rely on an LLM verifier that scores each rollout against rubrics. This introduces substantial training-time overhead, exposes optimization …

MCP-Atlas: A Large-Scale Benchmark for Tool-Use Competency with Real MCP Servers

Chaithanya Bandi, Razvan-Gabriel Dumitru, Ben Hertzberg, Divyansh Agarwal et autres

The Model Context Protocol (MCP) is emerging as a standard interface through which large language model (LLM) agents discover and invoke external tools. However, existing MCP evaluations fall short along three key axes: realistic multi-step workflows with cross-server orchestration, breadth across authentic …

CopySpec: Accelerating LLMs with Speculative Copy-and-Paste Without Compromising Quality

Razvan-Gabriel Dumitru, Minglai Yang, Vikas Yadav, Mihai Surdeanu

We introduce CopySpec, a simple yet effective technique to tackle the inefficiencies LLMs face when generating responses that closely resemble previous outputs or responses that can be verbatim extracted from context. CopySpec identifies repeated sequences in the model's chat history or context …

Variable Layerwise Quantization: A Simple and Effective Approach to Quantize LLMs

Razvan-Gabriel Dumitru, Vikas Yadav, Rishabh Maheshwary, Paul-Ioan Clotan et autres

We present a simple meta quantization approach that quantizes different layers of a large language model (LLM) at different bit levels, and is independent of the underlying quantization technique.Specifically, we quantize the most important layers to higher bit precision and less important …

us, it (code pays fourni par la source)

Relation-Aware Prompting Makes Large Language Models Effective Zero-shot Relation Extractors

Mahdi Rahimi, Razvan-Gabriel Dumitru, Mihai Surdeanu

While supervised relation extraction (RE) models have considerably advanced the state-of-theart, they often perform poorly in low-resource settings.Zero-shot RE is vital when annotations are not available either due to costs or time constraints.As a result, zero-shot RE has garnered interest in the …

us (code pays fourni par la source)

ConciseRL: Conciseness-Guided Reinforcement Learning for Efficient Reasoning Models

Razvan-Gabriel Dumitru, Darius Petelează, Vikas Yadav, Liangming Pan

Large language models excel at complex tasks by breaking down problems into structured reasoning steps.However, reasoning traces often extend beyond reaching a correct answer, causing wasted computation, reduced readability, and hallucinations.To address this, we introduce a novel hyperparameter-free conciseness score used as …

us (code pays fourni par la source)

CopySpec: Accelerating LLMs with Speculative Copy-and-Paste

Razvan-Gabriel Dumitru, Minglai Yang, Vikas Yadav, Mihai Surdeanu

We introduce CopySpec, a simple yet effective technique to tackle the inefficiencies LLMs face when generating responses that closely resemble previous outputs or responses that can be verbatim extracted from context.CopySpec identifies repeated sequences in the model's chat history or context and …

us (code pays fourni par la source)

Change Is the Only Constant: Dynamic LLM Slicing based on Layer Redundancy

Razvan-Gabriel Dumitru, Paul-Ioan Clotan, Vikas Yadav, Darius Petelează et autres

This paper introduces a novel model compression approach through dynamic layer-specific pruning in Large Language Models (LLMs), enhancing the traditional methodology established by SliceGPT. By transitioning from constant to dynamic slicing, our method leverages the newly proposed Layer Redundancy (LR) score, which …