ALT 2027 — 38th International Conference on Algorithmic Learning Theory

ALT 2027, the 38th International Conference on Algorithmic Learning Theory, runs 9-12 March 2027 in Leiden, Netherlands, organised by the Association for Algorithmic Learning Theory. ALT is the long-running venue where the mathematical guarantees behind machine learning are established, and the 2027 call for papers is aimed squarely at systems already in production: large language models and transformers, reinforcement learning including modern LLM post-training, robustness against training-data corruption, adversarial examples and LLM jailbreaks, learning under distribution shift and out-of-distribution generalization, and theoretical perspectives on trustworthy AI and AI safety covering privacy, adaptive data analysis, fairness and alignment. Papers were due 12 October 2026, with an author feedback period from 30 November to 6 December 2026 and notifications on 5 January 2027. Proceedings appear in the Proceedings of Machine Learning Research series.

• Paper submission deadlineOct 12, 2026
• Author feedback period opensNov 30, 2026
• Notification to authorsJan 5, 2027

Date

March 9–12, 2027

Location

Leiden, Netherlands

Format

In-Person

Event Type

Conference

Status

Upcoming

Pricing

Paid

Organizer

Association for Algorithmic Learning Theory (AALT)

Added

Sep 3, 2026

Updated

Sep 3, 2026

• ✓Machine learning researchers
• ✓AI research scientists
• ✓Heads of AI research
• ✓Applied scientists working on model evaluation, robustness and safety
• ✓PhD students in machine learning and statistics
• ✓Theoretical computer scientists and statisticians

If your organisation is being asked to certify that a deployed model is robust, private, fair or aligned, ALT is where the underlying guarantees are actually proved. A VP of AI or head of research gets four days of direct access to the people writing the theory behind LLM post-training, jailbreak robustness, distribution-shift detection and adaptive data analysis — the evidence base your risk, audit and regulatory functions will eventually demand, months or years before it reaches a vendor datasheet. It is also a concentrated recruiting venue: the authors presenting here are the researchers your competitors are hiring.

• ✓Call-for-papers topic area: large language models, transformers and related theoretical questions
• ✓Call-for-papers topic area: reinforcement learning, from classical control-theoretic perspectives to modern applications such as LLM post-training
• ✓Call-for-papers topic area: theoretical perspectives on trustworthy AI and AI safety, including privacy, adaptive data analysis, fairness and alignment
• ✓Call-for-papers topic area: robustness, including training-data corruption, adversarial examples and LLM jailbreaks
• ✓Call-for-papers topic area: learning under distribution shift, including domain adaptation and out-of-distribution generalization
• ✓Call-for-papers topic area: theoretical perspectives on deep learning — approximation, generalization and optimization, including implicit bias and overparameterization
• ✓Author feedback period, 30 November to 6 December 2026, ahead of notification on 5 January 2027

• ★The 2027 call for papers explicitly names large language models, transformers, LLM post-training and LLM jailbreaks — this is theory aimed at systems already deployed, not only at classical statistical learning.
• ★Trustworthy AI and AI safety are a first-class topic area, spanning privacy, adaptive data analysis, fairness and alignment — unusual for a learning-theory venue and directly relevant to anyone building an AI assurance case.
• ★A standing series with an established review process and PMLR proceedings: DBLP records the 36th edition in Milan (2025), the 35th in La Jolla (2024) and the 34th in Singapore (2023).
• ★Held in March in the Netherlands, filling the gap in the research calendar between NeurIPS in December and ICML in July.

Registration rates for ALT 2027 are not published on the official site as of September 2026. algorithmiclearningtheory.org currently carries only the call for papers and its submission, feedback and notification dates; registration details are expected nearer the March 2027 conference.

When is ALT 2027?

ALT 2027 takes place March 9–12, 2027 in Leiden, Netherlands.

Where is ALT 2027 held?

ALT 2027 is held in Leiden, Netherlands. The event format is in-person.

What is the ALT 2027 submission deadline?

The next ALT 2027 deadline is paper submission deadline on Oct 12, 2026. Remaining dates: Author feedback period opens — Nov 30, 2026; Notification to authors — Jan 5, 2027.

How much does ALT 2027 cost?

ALT 2027 is paid. Registration rates for ALT 2027 are not published on the official site as of September 2026. algorithmiclearningtheory.org currently carries only the call for papers and its submission, feedback and notification dates; registration details are expected nearer the March 2027 conference.

Who should attend ALT 2027?

ALT 2027 is aimed at Machine learning researchers, AI research scientists, Heads of AI research, Applied scientists working on model evaluation, robustness and safety, PhD students in machine learning and statistics, Theoretical computer scientists and statisticians.

What topics does ALT 2027 cover?

ALT 2027 covers Learning theory, Large language models and transformers, Reinforcement learning and LLM post-training, Trustworthy AI and AI safety, Robustness and adversarial examples, Distribution shift and out-of-distribution generalization, Deep learning theory, Optimization and overparameterization, Online learning and game theory, Privacy, fairness and alignment.

This page was written from 3 sources, 1 on domains other than algorithmiclearningtheory.org.

1. 1.algorithmiclearningtheory.org — alt2027vendor
2. 2.algorithmiclearningtheory.org — call for papersvendor
3. 3.dblp.org — index