76 citations · 112 across the 30 of their papers we have counts for
35 papers
GAW-PO: Preference Optimization with Gradient-Aligned Token Weights
Andreea Dutulescu, Stefan Ruseti, Mihai Masala +2
Most preference optimization methods, such as Direct Preference Optimization (DPO), apply preference supervision at the response level, although autoregressive language models are…
Large Language Models for Programming: Actually Fixing or Reimplementing Incorrect Code?
Alexandru Stefan Stoica, Traian Rebedea, Marian Cristian Mihaescu
Recent studies have shown that Large Language Models can effectively solve problems and fix bugs in diverse programming environments, including competitive programming. Existing ap…
Paved with True Intents: Intent-Aware Training Improves LLM Safety Classification Across Training Regimes
Jeremias Ferrao, Niclas Müller-Hof, Iustin Sîrbu +2
We argue that safety classifiers should model user intent as an explicit signal between the prompt and the final label. To study this, we introduce AIMS, a human-annotated dataset…
Defenses & Enablers For Skill Injection Attacks on Terminal Based Agents
Yoshinari Fujinuma, Varun Gangal, Traian Rebedea +4
Large language model (LLM) agents increasingly rely on reusable skills i.e. documents describing task-specific procedures. However, this introduces a new attack surface for agents…
"Înţelegi Româneşte?'' A Recipe for Romanian Vision-Language Models
Mihai Masala, Marius Leordeanu, Mihai Dascalu +1
Vision-Language Models (VLMs) largely follow the text-only LLM trajectory, excelling on English benchmarks but sharply degrading on low-resource languages, where neither large-scal…
SmellBench: Evaluating LLM Agents on Architectural Code Smell Repair
Ion George Dinu, Marian Cristian Mihăescu, Traian Rebedea
Architectural code smells erode software maintainability and are costly to repair manually, yet unlike localized bugs, they require cross-module reasoning about design intent that…
Training a General Purpose Automated Red Teaming Model
Aishwarya Padmakumar, Leon Derczynski, Traian Rebedea +1
Automated methods for red teaming LLMs are an important tool to identify LLM vulnerabilities that may not be covered in static benchmarks, allowing for more thorough probing. They…
Nemotron 3 Super: Open, Efficient Mixture-of-Experts Hybrid Mamba-Transformer Model for Agentic Reasoning
NVIDIA, :, Aakshita Chandiramani +544
We describe the pre-training, post-training, and quantization of Nemotron 3 Super, a 120 billion (active 12 billion) parameter hybrid Mamba-Attention Mixture-of-Experts model. Nemo…
Complexity-based code embeddings
Rares Folea, Radu Iacob, Emil Slusanschi +1
This paper presents a generic method for transforming the source code of various algorithms to numerical embeddings, by dynamically analysing the behaviour of computer programs aga…
Semi-Supervised Learning for Large Language Models Safety and Content Moderation
Eduard Stefan Dinuta, Iustin Sirbu, Traian Rebedea
Safety for Large Language Models (LLMs) has been an ongoing research focus since their emergence and is even more relevant nowadays with the increasing capacity of those models. Cu…
Pluralistic Behavior Suite: Stress-Testing Multi-Turn Adherence to Custom Behavioral Policies
Prasoon Varshney, Makesh Narsimhan Sreedhar, Liwei Jiang +2
Large language models (LLMs) are typically aligned to a universal set of safety and usage principles intended for broad public acceptability. Yet, real-world applications of LLMs o…
Improving Romanian LLM Pretraining Data using Diversity and Quality Filtering
Vlad Negoita, Mihai Masala, Traian Rebedea
Large Language Models (LLMs) have recently exploded in popularity, often matching or outperforming human abilities on many tasks. One of the key factors in training LLMs is the ava…
Meta-learning how to Share Credit among Macro-Actions
Ionel-Alexandru Hosu, Traian Rebedea, Razvan Pascanu
One proposed mechanism to improve exploration in reinforcement learning is through the use of macro-actions. Paradoxically though, in many scenarios the naive addition of macro-act…
MultiMatch: Multihead Consistency Regularization Matching for Semi-Supervised Text Classification
Iustin Sirbu, Robert-Adrian Popovici, Cornelia Caragea +2
We introduce MultiMatch, a novel semi-supervised learning (SSL) algorithm combining the paradigms of co-training and consistency regularization with pseudo-labeling. At its core, M…
Safety Through Reasoning: An Empirical Study of Reasoning Guardrail Models
Makesh Narsimhan Sreedhar, Traian Rebedea, Christopher Parisien
Reasoning-based language models have demonstrated strong performance across various domains, with the most notable gains seen in mathematical and coding tasks. Recent research has…
Aegis2.0: A Diverse AI Safety Dataset and Risks Taxonomy for Alignment of LLM Guardrails
Shaona Ghosh, Prasoon Varshney, Makesh Narsimhan Sreedhar +4
As Large Language Models (LLMs) and generative AI become increasingly widespread, concerns about content safety have grown in parallel. Currently, there is a clear lack of high-qua…
GIT-CXR: End-to-End Transformer for Chest X-Ray Report Generation
Iustin Sîrbu, Iulia-Renata Sîrbu, Jasmina Bogojeska +1
Medical imaging is crucial for diagnosing, monitoring, and treating medical conditions. The medical reports of radiology images are the primary medium through which medical profess…
Towards Inference-time Category-wise Safety Steering for Large Language Models
Amrita Bhattacharjee, Shaona Ghosh, Traian Rebedea +1
While large language models (LLMs) have seen unprecedented advancements in capabilities and applications across a variety of use-cases, safety alignment of these models is still an…
Unsupervised Extraction of Dialogue Policies from Conversations
Makesh Narsimhan Sreedhar, Traian Rebedea, Christopher Parisien
Dialogue policies play a crucial role in developing task-oriented dialogue systems, yet their development and maintenance are challenging and typically require substantial effort f…
"Vorbeşti Româneşte?" A Recipe to Train Powerful Romanian LLMs with English Instructions
Mihai Masala, Denis C. Ilie-Ablachim, Alexandru Dima +10
In recent years, Large Language Models (LLMs) have achieved almost human-like performance on various tasks. While some LLMs have been trained on multilingual data, most of the trai…
OpenLLM-Ro -- Technical Report on Open-source Romanian LLMs
Mihai Masala, Denis C. Ilie-Ablachim, Dragos Corlatescu +6
In recent years, Large Language Models (LLMs) have achieved almost human-like performance on various tasks. While some LLMs have been trained on multilingual data, most of the trai…
CantTalkAboutThis: Aligning Language Models to Stay on Topic in Dialogues
Makesh Narsimhan Sreedhar, Traian Rebedea, Shaona Ghosh +2
Recent advancements in instruction-tuning datasets have predominantly focused on specific tasks like mathematical or logical reasoning. There has been a notable gap in data designe…
Improving Legal Judgement Prediction in Romanian with Long Text Encoders
Mihai Masala, Traian Rebedea, Horia Velicu
In recent years,the entire field of Natural Language Processing (NLP) has enjoyed amazing novel results achieving almost human-like performance on a variety of tasks. Legal NLP dom…
NeMo Guardrails: A Toolkit for Controllable and Safe LLM Applications with Programmable Rails
Traian Rebedea, Razvan Dinu, Makesh Sreedhar +2
NeMo Guardrails is an open-source toolkit for easily adding programmable guardrails to LLM-based conversational systems. Guardrails (or rails for short) are a specific way of contr…