Semih Yavuz
@semih__yavuz
Followers
225
Following
200
Media
6
Statuses
58
Researcher at Salesforce. Interested in Deep learning and NLP. IMO medalist. Still love solving math puzzles.
Palo Alto
Joined September 2010
Super excited to share the SFR-embedding-2 model! 🎉 It has reclaimed the #1 spot on the MTEB leaderboard with significant improvements in several downstream tasks!🥇. Huge shoutout to the team:.@memray0 @YeLiu918 @mrnt0810 @JotyShafiq @yingbozhou_ai @CaimingXiong.
🎆I am pleased to announce the release of the latest version of the Salesforce Embedding Model (SFR-embedding-v2), which has reclaimed the top-1 position on the MTEB benchmark. ✨ Key Highlights:.🥇 Achieved the distinction of being the second model to surpass a 70+ performance
0
0
9
RT @SFResearch: 🆕Excited to announce SWERank, our code ranking framework for software issue localization. ➡️Paper: .
0
21
0
RT @SFResearch: 👁️ Looking for VLMs that go beyond generators to transform multimodal embeddings? Meet "VLM2VEC: Training Vision-Language….
0
28
0
RT @SFResearch: 🚀 AI is transforming industries, but how do we ensure it's delivering accurate, grounded answers? Introducing ContextualJud….
0
4
0
RT @CaimingXiong: As large language models increasingly serve as judges for evaluating other models during both development and deployment,….
0
17
0
RT @SFResearch: We built SFR-Embedding-Code to bridge a critical gap: . While text retrieval has advanced rapidly, code retrieval needed sp….
0
10
0
RT @Marktechpost: Salesforce AI Research Introduced CodeXEmbed (SFR-Embedding-Code): A Code Retrieval Model Family Achieving #1 Rank on CoI….
0
11
0
RT @SFResearch: 🚀 Introducing our latest breakthrough: the SFR-embedding model, a new champion on the MTEB benchmark! 🥇. But why does it ex….
0
5
0
RT @CaimingXiong: Introducing 🔥SFR-Embedding-Mistral🔥 has clinched the #1 spot on the MTEB leaderboard!🥇.Key highlights:.Retrieval and Rera….
huggingface.co
0
17
0
RT @CaimingXiong: We introduce 🔥XGen-7B 🔥, a new 7B LLM trained on up to 8K sequence length for 1.5T tokens. Achieves better or comparable….
github.com
Salesforce open-source LLMs with 8k sequence length. - salesforce/xgen
0
109
0
RT @Benioff: Salesforce 🔥XGen-7B 🔥new 7B LLM trained on up to 8K sequence length for 1.5T tokens. Achieves better or comparable results wit….
github.com
Salesforce open-source LLMs with 8k sequence length. - salesforce/xgen
0
50
0
RT @SFResearch: 🔥Introducing XGen-7B, a new 7B LLM trained on 8K seq. length for 1.5T tokens. Better or comparable results with MPT, Falcon….
github.com
Salesforce open-source LLMs with 8k sequence length. - salesforce/xgen
0
40
0
RT @SFResearch: Check out our #ACL2023 Accepted Papers! Congrats to all the authors! @silviocinguetta @CaimingXiong @alexfabbri4 @JiachengN….
0
15
0
RT @erenbali: Southeast Turkey was hit by a 7.8 magnitude earthquake that lasted 90 seconds. Over 100 buildings were reported to have colla….
0
746
0
RT @SFResearch: We are transforming the way we search in Salesforce, enabling your users to get actionable search results. Now, Einstein Se….
0
1
0
RT @SFResearch: 1/2: RnG-KBQA (Rank&Generate Approach for QA over Knowledge Bases) answers natural language questions over large-scale KBs,….
0
6
0
RT @SFResearch: Read our blog on #ACL2022. Congrats to all our authors for their accepted papers! ..
0
5
0
Excited to share our Rank-and-Generate (RnG-KBQA) approach sets new SoTA for knowledge base question answering. Paper, code, and models are released. Welcome to give it a try :) .Paper: Code:
github.com
Contribute to salesforce/rng-kbqa development by creating an account on GitHub.
Excited to share my internship work @SFResearch titled RnG-KBQA: Generation Augmented Iterative Ranking for Knowledge Base Question Answering. Paper: Code: Joint work with @semih__yavuz Kazuma Hashimoto @yingbozhou_ai @CaimingXiong
2
1
8
Joint work with @ShiyangLi6 @hassygo @VenaLi14 @mrnt0810 @nazneenrajani @xifengyan @yingbozhou_ai @CaimingXiong from @SFResearch . #NLP #ConversationalAI #DeepLearning #Generalizability #DataAugmentation #ICLR2021.
0
1
1
Prelim inspection of CoCo as data augmentation (CoCoAug) improves the accuracy of TripPy model from 55% to 60.5% on the original MultiWOZ evaluation set.
0
0
0