Faculty Profile: Prof. Graham Neubig — CMU LTI (NeuLab)
Faculty Profile: Prof. Graham Neubig — CMU LTI (NeuLab)
Position: Associate Professor（Full Professor track） Institution: Carnegie Mellon University, Language Technologies Institute Lab: NeuLab — https://www.cs.cmu.edu/~neulab/ Report date: 2026-06-12
Research Focus
Large language models, agents, code generation, small language models, LM reliability and evaluation, multilingual NLP. Organized CMU LLM Agent Workshop 2025.
Academic Profile
- PhD: Nara Institute of Science and Technology (Japan)
- Joined CMU LTI ~2016
- ACL, EMNLP, NAACL area chair; prolific senior author
- Organized CMU Agent Workshop (April 2025)
- Postdoc at Google and CMU background
Key Publications (2024–2025)
| Paper | Focus |
|---|---|
| OpenScholar / open research agents | Scientific RAG + agent for literature synthesis |
| Multilingual multimodal LLMs (EMNLP’25) | Grounding multilingual + multimodal LLMs |
| Various code generation, agent evaluation | NeuLab core output |
Verified Alumni Placement (frontier emphasis)
| Graduate | PhD Year | Placement | Role Family |
|---|---|---|---|
| Lucio Dery | 2024 | Google DeepMind | Research Scientist |
| Zhengbao Jiang | 2024 | Microsoft Research | Research Scientist |
| Shruti Rijhwani | 2022 | Google DeepMind | Research Scientist |
| Xinyi Wang | 2022 | Google DeepMind | Research Scientist |
| Pengcheng Yin | 2021 | Google DeepMind (Gemini code) | Research Scientist |
| Paul Michel | 2021 | Google DeepMind | Research Scientist |
| John Wieting | 2020 | Google DeepMind | Research Scientist |
| Chunting Zhou | 2022 | Meta AI | Research Scientist |
| Aditi Chaudhary | 2022 | Research Scientist | |
| Junxian He | 2022 | HKUST | Assistant Professor |
| Danish Pruthi | 2021 | IISc Bangalore | Assistant Professor |
| Junjie Hu | 2021 | UW-Madison | Assistant Professor |
Frontier gate: Google DeepMind ×7 — strongest of any advisor evaluated.
Fit with Weijia Zhang
| Dimension | Assessment |
|---|---|
| LLMs / NLP | ✅ Core |
| Agents / agentic AI | ✅ CMU Agent Workshop organizer, agent eval research |
| Code generation | ✅ Strong (NeuLab has code LLM work) |
| LM reliability / evaluation | ✅ Strong (NeuLab known for evaluation work) |
| RL post-training | ⚠️ Less direct, but adjacent via agent training |
| VLM / multimodal | ✅ Multilingual multimodal work |
| Frontier placement | ✅ Best verified record of all advisors evaluated |
Cold Email Strategy
- Reference: OpenManus-RL + GUIAgentDebugger as evidence of agent research experience
- Reference: MSRA experience (Microsoft Research connection — Neubig knows MSR well)
- Mention: Interest in agent evaluation, LM reliability, and the intersection of RAG and agents
Verdict
P1 套磁。 Best frontier placement record evaluated (DeepMind ×7). Well-established lab with consistent pipeline to top industry research roles. Research direction (LLMs + agents + evaluation) directly matches Weijia’s work. Lab is larger but highly productive.