Salesforce AI Research

Looking for the cutting-edge of AI research? Follow Salesforce AI Research to see how we're transforming enterprise technology through advanced innovations. From world models to agentic systems, discover the future of AI before it hits the market.

GIFT-Eval-CTX: An Open-Source Benchmark for Contextual Forecasting

🔗 Blog: sforce.co/4iEWybq
🔗 Paper: arxiv.org/abs/2610.17794

Energy operators, supply chain leaders, and IT teams often know what's coming, from a heatwave to a product launch. But that knowledge lives in text, and traditional forecasting models can only learn from historical numbers with no way to translate real-world context into quantitative adjustments.

GIFT-Eval-CTX measures how well LLMs and agents turn that forward-looking context into accurate numerical forecasts across 10,000 tasks from 20 real-world datasets.

→ Structured context about upcoming events reduced GPT-5.6 Sol's forecasting error by 40.7% and Claude Opus 4.8's by 29.6%
→ MoiraiAgent, which pairs an LLM with Moirai-2 and code execution, achieved 36.3% lower error than the best history-only model, with its lead growing on longer forecasts

The benchmark, dataset, and leaderboard are openly available for anyone building context-aware forecasting.

#EnterpriseAI #AgenticAI

Momentum might be the most underrated KPI in innovation.

The hardest call for most organizations? Knowing when to let a project go.

@Digitai and Linda A. Hill on what drives both. Watch on Salesforce+ → sforce.co/4zsYqe6

#EnterpriseAI

Interactive Visual Gym: Evaluating Visual Interaction Models Through Simulation

Can a visual AI assistant’s real-time advice actually help someone finish a procedural task? Meet Interactive Visual Gym (IVG), an open simulation framework for evaluating and training visual AI assistants in closed loop. We recreate cooking, household, and phone tasks in a simulation world alongside simulated users with different skills and preferences. These users act, make mistakes, ask questions, and respond to guidance – allowing us to test how accurately, appropriately, and promptly an assistant supports them as the activity unfolds.

→ 14 assistants, 315 scenarios: the best performing model, Gemini-4-Argon, completes only 53.4% of tasks on time
→ Human preferences match our simulator scores ~80% of the time
→ Training in the simulator lifts a 4B open model’s success rate’ from 33.3% to 42.2% on unseen tasks

Explore the paper, code, and demos: sforce.co/4zwQdWU

#EnterpriseAI #AgenticAI #FutureOfAI

Voice Models: What It Takes to Make Enterprise Voice Speak

Meet Resonance, our in-house family of voice models powering Salesforce and Slack:
→ Steno listens
→ Sonic converses and acts
→ Siren speaks back

Watch the full #DF26 session on Salesforce+: sforce.co/4dEz050

#EnterpriseAI