LM-KBC 2026

Deadlines
NLP/CORE Unranked

LM-KBC 2026

Workshop on Automated Knowledge Base Construction 2026 - LM-KBC - Shared Task

1793170800000Budapest, HungaryOfficial workshop site Link status unverified

LM-KBC 2026 is the 5th edition of the AKBC Shared Task, co-located with EMNLP 2026 in Budapest, focusing on predicting complete knowledge base entries using language models. Participants must generate accurate object strings for six relations given subject-relation pairs, under strict closed-book and parameter budget constraints, with evaluation based on string matching and numeric tolerance.

Official CFP Back to deadlines Verified September 9, 2026

Paper fit

Contribution paths

A strong submission should clearly identify its contribution and evaluate it appropriately.

System Paper

A 6-page ACL-style paper describing the system, submitted via OpenReview, including public code on GitHub.

Research areas in scope

01

Task Focus

Predicting complete knowledge base entries from language modelsClosed-book knowledge extractionString-based object predictionMulti-step and agentic inferenceNumeric and set-valued relation prediction

Policies worth checking twice

  • Single-blind review
  • Maximum 32B total neural parameters across all inference-time models
  • Quantization does not reduce counted parameter size
  • MoE models counted by total published parameters, not active ones
  • No web search, RAG, external corpora, or KB lookup allowed
  • No additional training, fine-tuning, or continued pretraining permitted
  • Non-neural components (e.g., rule-based filtering, deduplication) are allowed
  • Predictions must be submitted in JSONL format with ObjectEntities as strings

Official sources

Compiled from the official call for papers. The organizers’ pages remain authoritative.

Last verified September 9, 2026