SFR-VibeTrain: The Agent That Trains Agents A messaging-native agentic control plane for model and agent training. Describe the task, upload data (or don't), and it reasons, launches, analyzes, ablates, and deploys. sforce.co/4dwzYzo #FutureOfAI #EnterpriseAI #AgenticAI

A great moment for the lab! 🎉 "Don't Stop Early" earned its spot on the Industry Track for advancing enterprise deep research at scale. Controlled information flow and evidence-aware termination prevent premature stopping and uneven coverage. bit.ly/49fk2zQ #ACL2026

🎉 3 papers from @salesforce.com AI Research accepted to #ICML2026, spanning Mixture-of-Experts efficiency and multi-agent reasoning + verification. Learn more ↓

Our Infrastructure & DevOps team was featured by Google Cloud. Lavanya Karanam and Avinash Gudagi on how @salesforce.com is solving one of the hardest problems in large-scale AI training: keeping powerful GPUs fully fed. https://bit.ly/4n6HDc3 #FutureOfAI #EnterpriseAI

3 AI trends shaping #EnterpriseAI: simulation environments, ambient intelligence, and agent-to-agent communication. Itai Asseo on our focus on Enterprise General Intelligence (#EGI) in @gregorojstersek.bsky.social's #TDX26 recap. https://bit.ly/4mT4irW #FutureOfAI

(1/5) 📊 Don't Stop Early: Scalable Enterprise Deep Research with Controlled Information Flow and Evidence-Aware Termination bit.ly/49fk2zQ

Congratulations to our researches on winning an Outstanding Paper Award at #ICLR2026 for LLMs Get Lost in Multi-Turn Conversation! 🏆 Authors: Philippe Laban, Hiroaki Hayashi, Yingbo Zhou, Jennifer Neville https://bit.ly/3ZoC6T1

Attending #ICLR2026? Meet us at @salesforce.com#ICLR2026Booth #203 to hear the latest on how we're evolving enterprise general intelligence. For more, follow our journey at ICLR: https://bit.ly/4vLy9qohttps://bit.ly/4vLy9qo #FutureOfAI #EnterpriseAI

📍 Live from Rio: 21 papers at #ICLR2026, starting today. Our work spans agent architectures, reasoning, evaluation, deep research reliability, and scalable RL, tackling what matters most for #EnterpriseAI. https://sforce.co/4vRgqOs #FutureOfAI

Can AI coding assistants maintain effectiveness as codebases scale 100×? LoCoBench-Agent evaluates agents across 10K to 1M token contexts, spanning 8,000 scenarios in 10 programming languages and four difficulty tiers. https://sforce.co/4txCDj4 👥: Jielin Qiu & Huan Wang #EnterpriseAI #AgenticAI

The Illusion of Certainty: Decoupling Capability and Calibration in On-Policy Distillation: bit.ly/48iccVY On-policy distillation boosts accuracy but causes severe overconfidence. CaOPD uses a student-grounded empirical target for Pareto-optimal calibration. Code: bit.ly/4cUtCKO

At #TDX26, Itai Asseo @iiitaiii.bsky.social#TDX26on Enterprise General Intelligence: refining generic LLMs into reliable enterprise agents through AI Foundry and Agentforce Labs, including eVerse, the Learning Engine, and Agent Startup. https://sforce.co/3OrRYCb #FutureOfAI #EnterpriseAI

(1/7) We have 6 papers accepted to ACL 2026, advancing work across web agent evaluation, LLM reasoning verification, uncertainty quantification, long-context efficiency, and multilingual judge systems. ACL 2026 takes place July 2-7 in San Diego, California. #ACL2026 #FutureOfAI #EnterpriseAI

That's a wrap on #TDX26! The @salesforce.com#TDX26AI Research team was on the ground in San Francisco, connecting with the community and sharing the latest from our labs. From research demos to conversations about what's next in #EnterpriseAI, it was great to be part of the energy!

Reliability over raw power. 🔬 @aimagazine.bsky.social features Silvio Savarese on why enterprise AI value comes from integrated systems—not bigger models—and how AI Foundry is building that foundation. https://bit.ly/41z0G4E #SystemLevelAI #AIFoundry

AI Foundry: Turning foundational research into enterprise AI products faster. Silvio Savarese and Itai Asseo in TechFinitive: bit.ly/4ccYmFy #FutureOfAI #EnterpriseAI #AgenticAI

The model wars are over. Enterprise AI success now lives at the system level. Silvio Savarese and Itai Asseo in @technologymag.bsky.social: bit.ly/4ttSBu4 #FutureOfAI #EnterpriseAI #AgenticAI

Ambient intelligence is moving from research into live sales and service workflows. Silvio Savarese discusses with @cxtoday.com: bit.ly/47DUhJ1 #FutureOfAI #EnterpriseAI #AgenticAI

Three agentic AI trends shaping the enterprise through 2027. Silvio Savarese and Itai Asseo discuss AI Foundry with @cio.com: bit.ly/4sg6iMf #FutureOfAI #EnterpriseAI #AgenticAI

One demo. Reliable replay. No cloud calls. GPA turns a single recorded workflow into deterministic desktop automation, entirely on-device. 🔎 Explore GPA: bit.ly/48r7Onp 📖 Read the blog: sforce.co/4sYdhu8 #EnterpriseAI #GUIAutomation

Why #EnterpriseAI demands a shift from models to systems. Itai Asseo discusses AI Foundry with @diginomica.com: bit.ly/4scZwGT #FutureOfAI #AgenticAI

From One Demo to Reliable Automation: How GPA Reimagines GUI Process Automation https://sforce.co/4sYdhu8 Show it a workflow once. GPA replays it reliably, locally, and without brittle scripts to maintain. #FutureOfAI #EnterpriseAI #AgenticAI #GUIAutomation

(1/5) 🎙️ Building Enterprise Realtime Voice Agents from Scratch: A Technical Tutorial Paper: bit.ly/4seq7Ee 25+ open-source speech-to-speech models exist, but none shows how to build a complete streaming voice agent with function calling.

1/5 Enterprise Sales Copilot: Enabling Real-Time AI Support with Automatic Information Retrieval in Live Sales Calls bit.ly/4tamShg 🎙️ Reps lose 25–65 sec per query searching CRM systems mid-call. SalesCopilot fixes that.

Introducing GPA: GUI Process Automation Record one workflow demo. Replay it automatically — deterministic, fully local, and free. 100% success at ~10× the speed of Gemini 3 Pro's computer-use agent across 16 desktop tasks. 📝 Blog: salesforceairesearch.com/gpa 🔗 Paper: arxiv.org/abs/2604.01676

1/5 How Salesforce AI Research is Building Efficient RL Training for the Agentic Era Read full technical write-up: sforce.co/4bMqvUU

Multimodal LLMs for Human-AI Interaction: Foundations, Agents, and Inclusive Applications https://mllm4haii.github.io/ Tutorial T4 at EACL 2026 covers how MLLMs extend beyond text to reason across visuals and GUIs. 📅 Sun, March 29 | 9:00-12:30 GMT+1 #FutureOfAI #EnterpriseAI #EACL2026

(1/5) VoiceAgentRAG: Solving the RAG Latency Bottleneck in Real-Time Voice Agents Using Dual-Agent Architectures Paper: bit.ly/4sMmqp4 🗣️ Voice agents need sub-200ms latency, but vector DB queries alone take 50–300ms.

"Many of the old rulebooks simply don't apply anymore." — Itai Asseo, VP of Salesforce AI Research, on how AI Foundry connects research to real business problems through rapid customer collaboration. https://sforce.co/4rTPhao #FutureOfAI #EnterpriseAI

AI Foundry: Powering our big bets for research in 2026. https://sforce.co/4rTPhao #FutureOfAI #EnterpriseAI

"The problems that matter most for businesses don't live at the model level anymore. They live at the system level." — Silvio Savarese, EVP & Chief Scientist at @Salesforce https://sforce.co/4rTPhao #FutureOfAI #EnterpriseAI

Introducing AI Foundry: Accelerating System-Level #EnterpriseAI https://sforce.co/4rTPhao As frontier models mature, the hardest enterprise challenges live at the system level. #AIFoundry brings together research, customers, and partners to develop and validate new AI capabilities. #FutureOfAI

As AI moves toward multi-agent systems, the key challenge isn't just coordination — it's communication. Agents need their own semantic layer: shared protocols that let them negotiate and act on behalf of the organizations they represent.

Agents need more than natural language. They need a shared semantic layer to communicate and negotiate efficiently.

VIBEPASS: Can Vibe Coders Really Pass the Vibe Check? sforce.co/3NCuoT6 AI models score 90%+ on code generation, but under 50% on finding and fixing their own bugs. The gap between writing code and reasoning about faults remains wide. 🔍 #FutureOfAI

1/5 Agentic Confidence Calibration 📄 bit.ly/4kbjgs3 AI agents stay overconfident when they fail. HTC diagnoses the full execution trajectory, not just the final output.

[1/12] 🚨 Your AI isn't as good at coding as you think. Frontier models are crushing benchmarks. Demos look magical. But here’s the uncomfortable truth: 👉 They’re great at writing code 👉 They’re bad at knowing when it’s wrong Enter: VIBEPASS 🧵

AI agents can follow rules. But business negotiation runs on standards — contextual, interpretive, value-laden. That gap is the defining challenge ahead. Silvio Savarese and Sabastian Niles in @fortune.com https://bit.ly/4spc27e #EnterpriseAI #FutureOfAI

Enterprise General Intelligence (#EGI) > #AGI. Silvio Savarese on why solving enterprise corner cases is the real challenge 🎯

(1/5) Reference-guided LLM judges can bridge the gap between RLVR and RLHF 🧠 When given strong reference outputs, LLM judges become substantially more accurate—and can power effective self-improvement without external feedback. Paper: arxiv.org/abs/2602.16802

When two AI agents negotiate, helpfulness can backfire. They agree each other into absurdity. Our A2A Semantic Layer Framework turns that chaos into trusted, verifiable interaction. A2A Semantic Layer Blog: sforce.co/49GtDB8 #FutureOfAI #AIAgents

(🧵 1/5) InterviewSim: A Scalable Framework for Interview-Grounded Personality Simulation bit.ly/3P4cQj9 🔍 Can LLMs faithfully simulate real people? Most evaluations never check against what individuals actually said.

(🧵 1/5) MAS-ProVe — the first systematic study of process verification for multi-agent systems. Paper: arxiv.org/abs/2602.03053 Do multi-agent systems actually benefit from automatic process-level verification? This paper puts that question to a rigorous test.

(🧵 1/5) PTL (Prune–Tune Loop): compress LLMs to ~half their size without losing reasoning ability. Instead of pruning all at once, it iterates—small prune steps + lightweight recovery each round. 🧠 w/ @nuspress.bsky.social Paper: bit.ly/4qV4TtQ

Poisoning the Well: Search Agents Get Tricked by Maliciously Hosted Content sforce.co/40Z1Gi5 AI search agents can be fooled by planted fake content, ~80% of queries returned the attacker's answer in controlled tests. Authors: Shafiq Joty, Xuan Phi Nguyen, Shrey Pandit, Yifei Ming

Our work on "Echoing" was accepted at the Agents in the Wild workshop at ICLR 2026 When LLM agents interact without oversight, they can abandon their assigned roles — and standard metrics won't catch it. #ICLR2026 #FutureOfAI #EnterpriseAI

When LLM agents talk to each other, a new failure emerges: agents abandon their assigned roles and mimic their conversational partner instead. Rates reach as high as 70% with major models — and increased reasoning doesn’t help. 🧵 arxiv.org/abs/2511.09710

Our Singapore AI Research team marked their 7th anniversary this week with an FY27 kickoff — showcasing work across forecasting agents, enterprise insights, coding agents, and computer use agents. 🌏

Your MCP (Model Context Protocol) agent is reading 7,000 tokens to find 2. MCP+ fixes that. 🎯 mcp-plus.github.io

​​[1/6] MCP+: Precision Context Management for MCP Agents mcp-plus.github.io A post-processing layer that wraps your MCP clients and returns only what your agent needs — zero changes to your existing logic. 🎯