Kick-off Meeting TrustLLM

April 14, 10-11 (CET)

The project kick-off meeting will be in Linköping, Sweden, 20-21 November 2023. The event is for project consortium members and invited guests only.

Agenda

Day 1 (20 November)

08:30 Registrations, coffee
09:00 Welcome Fredrik Heintz & Dhafer Lahbib (PO)
09:15 Partner tour de table, 2 min each Everyone
09:45 Icebreaker
09:50 Break
10:20 Project overview Fredrik Heintz
10:40 WP2
11:00 WP3
11:20 WP4
11:40 WP5
12:00 WP6
12:20 Lunch at Universitetklubben (walking distance)
13:30 WP7
13:50 WP8
14:10 WP9
14:20 WP10
14:30 Break
15:00 WP Discussions
16:30 General Assembly
17:00 End of Day 1

18:30 Dinner at Stångs magasin, by the water in the city centre. We’re welcome at 18:30 for drinks, and dinner is at 19:00. Map: https://maps.app.goo.gl/pQjH9j9CfDnisByZ7

Day 2 (21 November)

09:00 Welcome Day 2 and summary discussions from day 1 (Fredrik Heintz)
09:30 Statement from the EU Project Officer (Dhafer Lahbib, EU Project Officer)
10:00 Break
10:30 HPC (Stefan Kesselheim)
11:00 Development infrastructure (Nico Flores-Herr)
11:30 Data (Amaru Cuba Gyllensten, AI Sweden)
12:00 13:20 Lunch at Universitetsklubben
13:20 13:50 Working together and finances (Trine Platou)
13:50 14:50 Pressing issues and next steps (Fredrik Heintz)
14:50 15:00 Concluding remarks (Fredrik Heintz)
15:00 End of meeting

Logistics

Venue

We will meet at the main campus of Linköping University, Campus Valla, in a room called Ada Lovelace situated in building B, entry 27. The address is Olaus Magnus väg, 583 30 Linköping, Sverige. The Google map link to the B 27 entrance is https://maps.app.goo.gl/jq1vmbGhBjUjd7gm9

Dinner is at Stångs magasin, by the water in the city centre. Map: https://maps.app.goo.gl/pQjH9j9CfDnisByZ7

Travel

The meeting starts at 9:00 on Monday 20 November and ends at 15:00 on Tuesday 21 November. We recommend travelling on Sunday evening and return to your home on Tuesday evening.

  • By train: Direct high-speed connection from Copenhagen, and train is also possible from Oslo. There is a direct night train from Hamburg.
  • By air: KLM flies from Amsterdam to Linköping. The KLM flight from Linköping arrives late on Sunday evening, and leaves on Tuesday at 16:30. It’s a tiny airport so you need maximum 1 h to check-in and “walk” to the gate (it’s crawling distance, really).  Leaving the meeting at 15:00 allows to catch this flight.
  • By air + train: Another possibility is to take the train to either Stockholm airport or Copenhagen airport (the direct train stops at the airport).

Busses are available from the city centre to the campus. Recommended bus stops are Nobeltorget or Universitetet. Bus number 4 from Trädgårdstorget or 3,4,12 from Resecentrum (the train station). Paying is easy, just blip your debit/credit card onboard. More information on Östgötatrafiken.

Hotels

We haven’t made any group bookings. Suggested hotels are Scandic Frimurarhotellet or Scandic City, but any hotel in the city centre will do. The conference dinner will be in the city centre.

N.B.: Linköping airport is very small. Leaving the campus at 15:00 will allow you to catch the 16:30 outgoing flight.

 

 

How can we make large language models more factually reliable? Can better data, external tools, and structured knowledge help reduce hallucinations? This TrustLLM webinar will focus on improving the factual trustworthiness of LLMs.

As large language models are increasingly used in real‑world and high‑stakes settings, their tendency to produce fluent but incorrect information remains a major challenge. This webinar presents three main contributions of the ongoing work from TrustLLM Work Package 3, which tackles factual reliability through three methodological approaches: data curation, tool learning, and structured knowledge extraction. Together, these three perspectives show how better data, external tool integration, and structured knowledge representations can jointly strengthen the factual reliability and trustworthiness of large language models.

The first topic introduces JQL (Judging Quality across Languages). JQL is a scalable method for curating high‑quality multilingual datasets by distilling LLM‑based annotations into lightweight models built on cross‑lingual embeddings. This approach demonstrates how systematic data curation across languages can directly improve the factual grounding of LLMs.
The second contribution explores how structured tool use can help anchor model outputs in real‑world information. Tool learning enables LLMs to interact with external systems—such as retrievers or specialized tools—allowing them to verify facts and reason over up‑to‑date sources rather than relying solely on internal representations.
Finally, we explore knowledge graph construction and ontology learning as a way to enhance factual consistency. By comparing single‑step and multi‑step reasoning strategies, this work investigates how LLMs can more reliably extract structured knowledge from text, supporting downstream reasoning and verification tasks.

Please note that this webinar will be recorded!

Register here

More from TrustLLM