AdrienBufort @Forbu14

hello :) changing weather forecasting at https://t.co/dULJXvwNYH France Joined July 2015

Tweets

236
Followers

57
Following

459
Likes

3K

AdrienBufort @Forbu14

5 days ago

grok code 1 is the best ratio (model perf / cost) you can find

0 0 3 48 0

Download Image

New paper 'Uncertainty Management in the Construction of Knowledge Graphs: A Survey' now available in TGDK! doi.org/10.4230/TGDK.3… This paper explores how to handle uncertainty/knowledge deltas in KG construction. #OrangeResearch Word done w/ @yoan_chabot and Miguel Couceiro

0 1 3 234 0

AdrienBufort @Forbu14

3 months ago

twitter what are the best discrete diffusion papers ?

0 0 0 43 0

Ricky T. Q. Chen @RickyTQChen

5 months ago

Links to papers: Generator Matching arxiv.org/abs/2410.20587 Discrete Flow Matching arxiv.org/abs/2412.03487 Adjoint Matching arxiv.org/abs/2409.08861 Adjoint Sampling arxiv.org/abs/2504.11713

1 5 17 1K 16

AdrienBufort @Forbu14

6 months ago

trying @lovable_dev today ! very good product (on part with replit)

0 0 1 79 0

elio @elyxlz

7 months ago

just realized why deepseek did all of this ... doing RL fine-tuning on the stock market with returns as a reward and historical data + news as input isn't as far fetched as you think it is.

81 65 2K 202K 795

Yoan Chabot @yoan_chabot

a year ago

How to detect complex anomalies in #network infrastructures? At @orange, we believe that graph representation is key! Discover NORIA, a supervision solution based on #KnowledgeGraphs #AI 🔗hellofuture.orange.com/en/noria-netwo… @yoan_chabot @TailhardatL @rtroncy @orange_Future @EURECOM 1/n

1 7 11 946 0

Download Image

Barboule Camille @BarbouleCamille

2 years ago

Excited to share our research on domain-specific #LLM adaptation at #GDRTAL next week! Tailoring LLMs to company tech insights boosts value. We'll compare Full/#LoRA/#QLoRA ft methods, seeking best perf/cost balance. @Orange_Future @vp_huynh @Forbu14 @yoan_chabot @ghpu @tau_labbe

0 3 6 676 0

Steve Jarrett @stevejarrett

2 years ago

The open source @Meta #LLaMA #AI will allow us to create much smaller, faster, and more energy efficient yet still powerful models. At @orange we are committed to responsible AI that gives superpowers to our employees, our customer interactions, and our networks. We are impressed…

0 9 28 4K 2

Steve Jarrett @stevejarrett

3 years ago

Very clear thread on the expected business value from #PredictiveAI vs #GenerativeAI over the next few years from @SnorkelAI CEO Alex Ratner. High quality data will continue to be the bottleneck for achieving high quality AI model outputs…and high ROI.

Alex Ratner @ajratner

3 years ago

19 103 485 161K 254

Download Image

0 8 29 5K 6

Stéphane Richard @srichard

4 years ago

Agression inqualifiable chez nos confrères de @bouyguestelecom aujourd’hui. Pensées attristées aux proches des victimes et solidarité avec leurs collègues. Les incivilités dans nos boutiques se multiplient et ce drame est un terrible signal d’alarme.

42 294 955 0 2

ML Review 💙💛 @ml_review

4 years ago

"Applications of Deep Neural Networks" [569 pages] 📚by @jeffheaton pdf arxiv.org/pdf/2009.05673… abs arxiv.org/pdf/2009.05673…

4 86 365 0 157

Download Image

Papers with Code @paperswithcode

4 years ago

💥 Deep learning makes strides on tabular data! In this week’s newsletter, we summarize recent developments and papers using deep learning models for tabular data... and much more. Read on below: paperswithcode.com/newsletter/12

0 127 527 0 135

Download Image

Papers with Datasets @paperswithdata

4 years ago

🐙 GitTables: A corpus of currently 1.7M relational tables extracted from CSV files in GitHub. Table columns in GitTables have been annotated with more than 2K different semantic types from Schema.org and DBpedia. paperswithcode.com/dataset/gittab…

0 37 176 0 41

Download Image

Aran Komatsuzaki @arankomatsuzaki

4 years ago

SegFormer: Simple and Efficient Design for Semantic Segmentation with Transformers Proposes SegFormer, a simple, efficient yet powerful semantic segmentation framework which unifies Transformers with MLP decoders. abs: arxiv.org/abs/2105.15203 code: github.com/NVlabs/SegForm…

3 22 107 0 20

Download Image

Aran Komatsuzaki @arankomatsuzaki

4 years ago

Self-Attention Between Datapoints: Going Beyond Individual Input-Output Pairs in Deep Learning Introduces a general-purpose deep learning architecture that takes as input the entire dataset instead of processing one datapoint at a time. arxiv.org/abs/2106.02584

7 113 495 0 140

Download Image

Aran Komatsuzaki @arankomatsuzaki

4 years ago

Pretraining Representations for Data-Efficient Reinforcement Learning Proposes SGI, which significantly surpasses prior work on Atari with the steps limited to 100k with an improved unsupervised goal-conditioned RL. abs: arxiv.org/abs/2106.04799 code: github.com/mila-iqia/SGI