@SAYg_7

Senior Research Scientist @ Megagon Labs, US. My research interests include RAG and function calling in LLMs. 七帝柔道を応援しています。

seijimaekawa.github.io/
Joined June 2014
  • Tweets331
  • Following187
  • Followers280
24Media
ラジオ番組に出ました。会社辞めて、博士課程に入って、海外就職することについて話しました。今、修士・博士学生の方の役に立つと嬉しいです。 youtu.be/Hhl-hS_qC2c

営業から博士に戻り海外に就職する公聴会直前の心持ち：情処ラジオ

ダイバーシティを考える情報処理学会の情処ラジオでは、いろんな方の声を聞いています。※動画の場合字幕は自動でつけており、修正していませ...

youtube.com
3
14
5,709
My poster session is starting! Stop at #284 if you are in #EMNLP
22
611
来週 EMNLP で現地発表します！興味のある方ぜひポスター発表 (10/25 11:00~) もしくはMegagon Labsのスポンサーブースにお越しください！ PhD学生のインターンを募集していますので、LLM・RAG・Visual Understandingに興味のある方はぜひ声かけてください！ paper: arxiv.org/abs/2604.03403 #EMNLP

Align Then Adapt: Label-Efficient Adapter Learning for Asymmetric...

Dense retrieval systems increasingly face an asymmetry between complex instruction-like queries and relatively simple, static document collections. While stronger embedders can better understand...

arxiv.org
2
2
15
560
発表では、二つの異なる埋込モデルの潜在空間を線形層で繋ぎ合わせ、弱い方のモデルから検索精度を大幅に向上できることを話します。 例えば、新しい埋込モデルが出た時、corpusのembeddingは古いままでquery側に新モデル+線形層という構築で検索精度アップが狙えます。少量の学習データでも使えます！
97
Happy to share that our paper has been accepted to AACL-IJCNLP findings 🎉🎉 Thank you to the first author @nimashahbazi90 paper: arxiv.org/abs/2604.02444

From Textual Columns to Query Plans: A Unified Relational-Semantic...

Real-world table question answering often involves hybrid schemas in which some query-relevant information is explicit in relational columns, while other attributes, predicates, or join conditions...

arxiv.org
1
2
16
1,141
Seiji Maekawa retweeted
Seiji Maekawa@SAYg_7Sep 13
Megagon Labs ではインターンを募集しています！ シリコンバレーでRAGやTool Calling、Vision&Text Understanding に関する研究をしたい学生の方、ぜひ検討してみてください。相談したいことがあればぜひご連絡ください！
Megagon Labs
@MegagonLabsSep 12
We’re hiring a Fall/Winter PhD Research Intern at Megagon Labs. We’re looking for candidates with research or technical experience in any one or more of the following areas: • Retrieval-Augmented Generation (RAG) • Function Calling • Document Understanding / PDF Parsing You do not need experience across all three areas. Strong experience in any one of these topics is relevant. Interested in joining us this fall? Learn more and apply → megagon.ai/careers/internshi… Applicants must have or be able to obtain appropriate U.S. work authorization for the internship period. #AI #NLP #LLMs #internships #phdposition
6
16
3,471
Megagon Labs ではインターンを募集しています！ シリコンバレーでRAGやTool Calling、Vision&Text Understanding に関する研究をしたい学生の方、ぜひ検討してみてください。相談したいことがあればぜひご連絡ください！
Megagon Labs
@MegagonLabsSep 12
We’re hiring a Fall/Winter PhD Research Intern at Megagon Labs. We’re looking for candidates with research or technical experience in any one or more of the following areas: • Retrieval-Augmented Generation (RAG) • Function Calling • Document Understanding / PDF Parsing You do not need experience across all three areas. Strong experience in any one of these topics is relevant. Interested in joining us this fall? Learn more and apply → megagon.ai/careers/internshi… Applicants must have or be able to obtain appropriate U.S. work authorization for the internship period. #AI #NLP #LLMs #internships #phdposition
6
16
3,471
Seiji Maekawa retweeted
Seiji Maekawa@SAYg_7Aug 27
Happy to share that our paper has been accepted to EMNLP2026 main 🎉🎉 Huge kudos to co-authors!! Moin, @PPezeshkpour @estevamhruschka Link: arxiv.org/abs/2604.03403 The quick summary is ↓↓↓

Align Then Adapt: Label-Efficient Adapter Learning for Asymmetric...

Dense retrieval systems increasingly face an asymmetry between complex instruction-like queries and relatively simple, static document collections. While stronger embedders can better understand...

arxiv.org
1
3
31
1,780
Happy to share that our paper has been accepted to EMNLP2026 main 🎉🎉 Huge kudos to co-authors!! Moin, @PPezeshkpour @estevamhruschka Link: arxiv.org/abs/2604.03403 The quick summary is ↓↓↓

Align Then Adapt: Label-Efficient Adapter Learning for Asymmetric...

Dense retrieval systems increasingly face an asymmetry between complex instruction-like queries and relatively simple, static document collections. While stronger embedders can better understand...

arxiv.org
1
3
31
1,780
Your vector index is huge and costly to rebuild, but your users’ queries keep getting more complex. We show how to pair that fixed index with a stronger query encoder using a lightweight adapter and <100 labels/task—without re-indexing.
163
QUINTET.6 面白かった。ただ武徳殿は冷房ないのでめちゃくちゃキツそう
1
195
何年経っても後輩の七大戦の試合観るのは緊張する。画面越しでも
1
299

博士に進むか、就職するか。甥っ子との会話で見えてきた「本当の悩み」

Article · @miraibouken
1
36
21
228
107,708
Seiji Maekawa retweeted
Megagon Labs
@MegagonLabsApr 23
📢 Attending #ICLR 2026? @EstevamHruschka, Lab Director and Staff Researcher, will be presenting Towards Reliable Benchmarking: A Contamination Free, Controllable Evaluation Framework for Multi-step #LLM Function Calling. Not attending? Read the paper here: megagon.ai/publications/fram… Paper by @SAYg_7, Jackson Hassell, @PPezeshkpour, @tommmitchell, @estevamhruschka
1
4
9
625
Seiji Maekawa retweeted
Megagon Labs
@MegagonLabsApr 23
Attending #ICLR 2026 in Brazil? Learn how data representation affects LLM response, with a poster presentation on 𝗦𝗮𝗺𝗲 𝗖𝗼𝗻𝘁𝗲𝗻𝘁, 𝗗𝗶𝗳𝗳𝗲𝗿𝗲𝗻𝘁 𝗥𝗲𝗽𝗿𝗲𝘀𝗲𝗻𝘁𝗮𝘁𝗶𝗼𝗻𝘀: 𝗔 𝗖𝗼𝗻𝘁𝗿𝗼𝗹𝗹𝗲𝗱 𝗦𝘁𝘂𝗱𝘆 𝗳𝗼𝗿 𝗧𝗮𝗯𝗹𝗲 𝗤𝗔 with @estevamhruschka Not attending? Read the paper here: megagon.ai/publications/same… Paper by Yue Zhang, @SAYg_7, @nikita_bhutani #AI #LLMs #Research #PhD
2
3
351
今年はICLRに行けませんが、他のMegagonメンバーが2本の論文を発表してくれます！ ①ツール呼出の連鎖が長くなるとLLMがどう間違えるかを人工データで検証→途中経過のおさらいが重要と判明。 ②内容が同じ綺麗な表とテキスト混じりの表（仕事でよくある）に対するLLMの返答の変化を検証。 リンク↓↓
1
1
19
2,351
ツール呼出の論文：arxiv.org/abs/2509.26553 データ表現の違いの論文：arxiv.org/abs/2509.22983 主著と自分がメインメンターをしたインターンによる論文です。全ての共著者の方々に感謝です！

Towards Reliable Benchmarking: A Contamination Free, Controllable...

Existing benchmarks for tool-augmented language models (TaLMs) lack fine-grained control over task difficulty and remain vulnerable to data contamination. We present FuncBenchGen, a unified,...

arxiv.org
1
239
2026年、嵩賞（大阪大学卒業の若手研究者賞）をいただきました！ 博士研究でお世話になった先生・関係者の皆様、そしてMegagonでの同僚たちに非常に感謝しています！
2
2
27
1,658
Two papers have been accepted to #ICLR2026 🥳 (1 x first-author, 1 x primary mentor) Huge kudos to all the co-authors!
1
48
2,278
📕 “Efficient Context Selection for Long-Context QA: No Tuning, No Iteration, Just Adaptive‑k” Oral presentation @ Question Answering 11/5 Wed. 15:15 📄arxiv: arxiv.org/abs/2506.08479 ⚙️code: github.com/megagonlabs/adapt…

Efficient Context Selection for Long-Context QA: No Tuning, No...

Retrieval-augmented generation (RAG) and long-context language models (LCLMs) both address context limitations of LLMs in open-domain question answering (QA). However, optimal external context to...

arxiv.org
1
1
337
Megagon Labs ではインターンを募集しています！ シリコンバレーでNLP/LLMsや複合AI Systemsの研究をしてみたい学生の方、ぜひ検討してみてください。期間や時期等、相談したいことがあればぜひ連絡ください！
Megagon Labs
@MegagonLabs2 Oct 2025
📢 Internship applications are now open for the Summer 2026 term. Join Megagon Labs to research: 💠LLMs & NLP 💠Human-centered AI 💠Data AI 💠Compound AI Systems Interns at Megagon Labs work alongside our research teams, advancing projects that bridge academic research and real-world impact. Read about their experiences: megagon.ai/2025-internship-e… 👉 Apply here: megagon.ai/careers/ #AI #NLP #Internships #LLMs
4
25
5,316
Load more