ServiceNow AI Research

1

GitChameleon: Evaluating AI Code Generation Against Python Library Version Incompatibilities
GitChameleon: Evaluating AI Code Generation Against Python Library Version Incompatibilities
The rapid evolution of software libraries presents a significant challenge for code generation models, which must adapt to frequent …
Annual Meeting of the Association for Computational Linguistics (ACL),  2026.
PDF Cite
Augmenting LLM Reasoning with Dynamic Notes Writing for Complex QA
Augmenting LLM Reasoning with Dynamic Notes Writing for Complex QA
Language Resources and Evaluation Conference,  2026.
PDF Cite
DRBench: A Realistic Benchmark for Enterprise Deep Research
DRBench: A Realistic Benchmark for Enterprise Deep Research
We introduce DRBench, a benchmark for evaluating AI agents on complex, open-ended deep research tasks in enterprise settings. Unlike …
International Conference on Learning Representations,  2026.
PDF Cite
Grounding Computer Use Agents on Human Demonstrations
Grounding Computer Use Agents on Human Demonstrations
Building reliable computer-use agents requires grounding: accurately connecting natural language instructions to the correct on-screen …
International Conference on Learning Representations,  2026.
PDF Cite
No, of Course I Can! Deeper Fine-Tuning Attacks That Bypass Token-Level Safety Mechanisms
No, of Course I Can! Deeper Fine-Tuning Attacks That Bypass Token-Level Safety Mechanisms
Leading language model (LM) providers like OpenAI and Anthropic allow customers to fine-tune frontier LMs for specific use cases. To …
International Conference on Learning Representations,  2026.
PDF Cite
StarFlow: Generating Structured Workflow Outputs From Sketch Images
StarFlow: Generating Structured Workflow Outputs From Sketch Images
Workflows are a fundamental component of automation in enterprise platforms, enabling the orchestration of tasks, data processing, and …
European Chapter of the Association for Computational Linguistics (EACL),  2026.
PDF Cite Video
AlignVLM: Bridging Vision and Language Latent Spaces for Multimodal Document Understanding
AlignVLM: Bridging Vision and Language Latent Spaces for Multimodal Document Understanding
Aligning visual features with language embeddings is a key challenge in vision-language models (VLMs). The performance of such models …
Neural Information Processing Systems (NeurIPS),  2025.
PDF Cite Video
Causal Differentiating Concepts: Interpreting LM Behavior via Causal Representation Learning
Causal Differentiating Concepts: Interpreting LM Behavior via Causal Representation Learning
Language model activations entangle concepts that mediate their behavior, making it difficult to interpret these factors, which has …
Neural Information Processing Systems (NeurIPS),  2025.
PDF Cite
How to Train Your LLM Web Agent: A Statistical Diagnosis
How to Train Your LLM Web Agent: A Statistical Diagnosis

Large language model (LLM) agents for web interfaces have advanced rapidly, yet open-source systems still lag behind proprietary …

Neural Information Processing Systems (NeurIPS),  2025.
PDF Cite
Rendering-Aware Reinforcement Learning for Vector Graphics Generation
Rendering-Aware Reinforcement Learning for Vector Graphics Generation
Scalable Vector Graphics (SVG) offer a powerful format for representing visual designs as interpretable code. Recent advances in …
Neural Information Processing Systems (NeurIPS),  2025.
PDF Cite