Unlock work experiences of the future. Join @ServiceNowRSRCH as we advance the state-of-the-art in Enterprise AI. #ServiceNowResearch #LifeAtNow #Hiring
Joined July 2016
- Tweets 2,332
- Following 1,253
- Followers 18,671
- Likes 4,198
460 Photos and videos
๐ Grok Voice Think Fast 1.0 (@xAI) lands on the Pareto frontier on EVA-Bench โ no system in the eval beats it on accuracy without sacrificing experience, or vice versa.
๐ Leaderboard: servicenow.github.io/eva/#reโฆ
@elonmusk #VoiceAgents #ServiceNowResearch #EVABenchย #GrokVoice #xAI
4
25
96
114,260
โญ ๐๐ฉ๐-๐๐ฒ๐ป๐ฐ๐ต ๐๐ฎ๐๐ฎ ๐ฎ.๐ฌ: ๐ฏ ๐๐ผ๐บ๐ฎ๐ถ๐ป๐, ๐ญ๐ฎ๐ญ ๐ง๐ผ๐ผ๐น๐, ๐ฎ๐ญ๐ฏ ๐ฆ๐ฐ๐ฒ๐ป๐ฎ๐ฟ๐ถ๐ผ๐
We just published an article detailing the major expansion we have done to the data behind EVA-Bench.
๐๏ธ Data: huggingface.co/datasets/Servโฆ
๐ Article: huggingface.co/blog/ServiceNโฆ
#VoiceAgents #OpenSource #Data #AIResearch #ServiceNowResearch
8
13
654
Thrilled to see this benchmark, EnterpriseOps-Gym, featured in Jensen Huang's Computex keynote โ a testament to the Core AI team's work building rigorous agentic evaluation for the enterprise. ๐
Jun 2
Big moment for the whole ๐๐ป๐๐ฒ๐ฟ๐ฝ๐ฟ๐ถ๐๐ฒ๐ข๐ฝ๐-๐๐๐บ team. ๐
๐๐ฒ๐ป๐๐ฒ๐ป announced Apriel on the K25 stage. This year, EnterpriseOps-Gym made it into his ๐๐ผ๐บ๐ฝ๐๐๐ฒ๐
๐ธ๐ฒ๐๐ป๐ผ๐๐ฒ. ๐
NVIDIA announced ๐ก๐ฒ๐บ๐ผ๐๐ฟ๐ผ๐ป ๐ฏ ๐จ๐น๐๐ฟ๐ฎ, the largest Nemotron 3 model to date, with 550B parameters / 55B active ..and it was incredibly exciting to see results on our benchmark featured in the keynote.
Enterprise agents need benchmarks that go beyond static QA, ones that test:
- long-horizon planning
- reliable tool use
- stateful workflows
- enterprise realism and ability to act without breaking things downstream
That is exactly why we built EnterpriseOps-Gym. Huge congrats to the Nemotron team for building agentic capability at the frontier level of intelligence and cost. More to come. ๐ฅ @NVIDIAAI @nvidia @ServiceNowRSRCH @ServiceNowNews @sagardavasam @shiva_malay @PShravannayak
enterpriseops-gym.github.io/
7
853
ServiceNow AI Research retweeted
MosaicLeaks is now on arXiv. The Mosaic Effect captures a simple idea: small fragments can look harmless alone, but become revealing in aggregate. Deep research agents can leak enterprise information in exactly this way.
1/9
1
2
6
1,200
๐ ACM CAIS 2026 Best Paper Award!
"Malice in Agentland: Down the Rabbit Hole of Backdoors in the AI Supply Chain" was honored at the first conference dedicated to agentic systems ๐
Read the story ๐ servicenow.com/in/workflow/aโฆ
1
1
4
545
๐ Full paper: arxiv.org/pdf/2510.05159
Huge congrats to the team: @LeoBoisvert, @AbhayPuri98, Chandra Kiran Reddy Evuru, Nazanin Sepahvan, @alexandredrouin, @alex_lacoste_, @NicolasChapados, Quentin Cappart, @jstanl, @DjDvij ๐
2
1
367
Proud to share the latest from our Core AI colleagues! โ rethinking what world models should (and shouldn't) be for enterprise agents. ๐
May 13
๐จThe World Model Myth in Enterprise Systems๐จ
Everyoneโs obsessed with world models for agents.
But in enterprise systems? The entire "world" (rules, workflows, transitions) is readable at runtime. No hidden dynamics.
We asked the dangerous question:
Do enterprise systems need learned world models?
1
1
7
1,423
ServiceNow AI Research retweeted
Our first vLLM V0โV1 run on PipelineRL looked broken.
@ehsk0 and I almost reached for an objective-side correction. That would have been the wrong fix.
The real problem: four mismatches in the rollout backend.
๐งต
1
4
7
2,489
ServiceNow AI Research retweeted
Boris Cherny, Head of Claude Code, on the importance of ServiceNow $NOW
95
44
550
88,379
ServiceNow AI Research is heading soon to Seoul with 4 papers accepted at #ICML2026! ๐
Our work spans retrieval, multimodal forecasting, knowledge distillation, and diffusion language models.
Congrats to all the researchers involved! ๐
1
1
14
2,396
Here are the 4 papers accepted at #ICML2026! ๐๐ฐ๐ท
๐ Hierarchical Retrieval at Scale
๐ Overcoming the Modality Gap in Context-Aided Forecasting
๐ Privileged Information Distillation for Language Models
๐ DiffuMamba: High-Throughput Diffusion LMs with Mamba Backbone
See you in Seoul!
icml.cc/
1
3
1,518
(1/8)
๐ Introducing Super Apriel: One Checkpoint, Many Speeds
Train once โ serve at any speed-quality tradeoff
We release:
โ 15B supernets with 4 mixers/layer
โ Training code (Fast-LLM)
โ vLLM serving extension
๐งต How it works โ
1
10
32
10,553
(7/8)
๐ Bonus: speculative decoding for free
Use all-FA as verifier, fast preset as draft
Same checkpoint โ up to 2.75ร speedup, no quality loss
๐ Paper: arxiv.org/pdf/2604.19877
๐ค Models: huggingface.co/ServiceNow-AIโฆ
๐ป Code: github.com/ServiceNow/Fast-Lโฆ
1
1
4
1,560
(8/8)
๐ Amazing team that made this possible:
@ostap__alex @Ray97369304 @tscholak @alirezamh_ Aman Tomar Shruthan Radhakrishna Denis Kochetkov Joel Lamy-Poirier @nandahkrishna @muchomuchacho @SathwikTejaswi Srinivas Sunkara @vbecaert
1
7
1,458
ICLR 2026 is underway in Rio de Janeiro ๐
The ServiceNow AI Research team has 7 papers accepted across the main conference, workshops, and blogpost track โ covering AI safety, enterprise benchmarking, computer-use agents, and more.
Attending ICLR? Go catch their sessions! ๐
#ICLR2026 #AIResearch #MachineLearning
1
2
1,592
๐ No, Of Course I Can! โ Deeper Fine-Tuning Attacks That Bypass Token-Level Safety Mechanisms
๐ DRBench: A Realistic Benchmark for Enterprise Deep Research
๐ Grounding Computer Use Agents on Human Demonstrations
๐ Hierarchical Retrieval at Scale: Bridging Transparency and Efficiency โ Workshop
๐ Overcoming the Modality Gap in Context-Aided Forecasting โ Workshop
๐ CUA-Suite: Expert Trajectories and Pixel-Precise Grounding for Computer-use Agents โ Workshop
๐ Destruction is a General Strategy to Learn Generation โ Blogpost track
1,222