Menu

MLRC 2026: Announcing the Accepted Papers

We are delighted to announce the 16 papers accepted to the Machine Learning Reproducibility Challenge (MLRC) 2026, an official NeurIPS 2026 track. The papers will be presented in person at NeurIPS 2026 in Sydney, Australia (December 6–13, 2026), alongside papers from the Main Track and the Evaluations & Datasets Track. Each paper was first accepted at the Transactions on Machine Learning Research (TMLR), and subsequently underwent a compatibility review by the MLRC committee to confirm its suitability for the track.

Machine learning advances on the strength of empirical claims, yet the field’s capacity to verify those claims has not kept pace with its capacity to produce them. Independent reproductions are seldom published and rarely credited. MLRC exists to close this gap.  Its studies re-examine central claims, establish which results replicate, and test how far the claims generalize beyond their original setting, building a public, peer-reviewed record of what the field actually knows.

This year’s papers examine results in mechanistic interpretability, vision-language models, language model sampling, offline reinforcement learning, fairness, and agent evaluation. Two independent studies revisit the same work on bilinear MLPs, and one paper turns to NeurIPS itself, auditing the evidence trace behind its 2025 contributions.

Accepted Papers from the Transactions on Machine Learning Research (TMLR)

The MLRC poster session will be held on Thursday, December 10, 2026, 5:00-8:00 PM Sydney time. We invite all attendees to visit the posters and meet the authors.

  1. A Reproduction Study of Weight-Based Mechanistic Interpretability in Bilinear MLPs

Itay Erlich, May Ben Zion, Jacob Shashoua

https://openreview.net/forum?id=6k7qRdz7rD

  1. Auditing Closed-Loop Learning in Recurrent Neural Networks: Reproduction, Robustness, and Generalization

Aarav Sinha

https://openreview.net/forum?id=e0pSqIqwXO

  1. [Re] Boosting the Visual Interpretability of CLIP via Adversarial Fine-Tuning

Anton Nuzhdin, Andrei Gruzitski, Balázs Egyed, Aleksandr Raudvee

https://openreview.net/forum?id=uPXRBZfkYy

  1. Evaluating the evidence trace of NeurIPS 2025 contributions with agentic code auditing

Pascal Iversen, Ferdous Nasri, Bernhard Y. Renard, Katharina Baum

https://openreview.net/forum?id=xjCJY8Xxcm

  1. [Re] FairDICE: A Fair Tradeoff in Multi-objective Offline RL

Peter Adema, Karim Galliamov, Aleksey Evstratovskiy, Ross Geurts

https://openreview.net/forum?id=Tr6MBt0hAj

  1. Min-p, Max Exaggeration: A Critical Analysis of Min-p Sampling in Language Models

Rylan Schaeffer, Joshua Kazdan, Yegor Denisov-Blanch

https://openreview.net/forum?id=rCupWGkP6W

  1. Opening Up a New Layer: A Deeper Look into “Interpreting CLIP with Hierarchical Sparse Autoencoders”

Nora Kahleogullari, Ada Romanowski, Mathan Sundarrajan, Sally Veerman

https://openreview.net/forum?id=l0qZkSk2aH

  1. Reproducibility study of “Bilinear MLPs enable weight-based mechanistic interpretability”

Hamid Reza Ahmadi, Mihnea Andrei Bârsan, Jan Jelínek, Robert-Ștefan Sofroni

https://openreview.net/forum?id=Bnx3hf3mWp

  1. Reproducibility study of FACTER: Fairness-Aware Conformal Thresholding and Prompt Engineering

Elena Belkina, Alexandra Elena Antonica, Brigitta Varga, Mattia Sferrazza, Clara Rus

https://openreview.net/forum?id=vX5mPaLdeQ

  1. Reproducing FACTER: Fairness via Conformal Thresholding and Prompt Repair

Oscar Miró López-Feliu, Daimy van Loo, Xanthos Kekkos, Mikel Blom, Clara Rus

https://openreview.net/forum?id=4BPFVex4EM

  1. Revisiting “Edit Away and My Face Will not Stay: Personal Biometric Defense against Malicious Generative Editing”

Luis Vitor Zerkowski, Soham Chaudhuri, Finley Helms, Jelle Sombekke, Udit Thakur

https://openreview.net/forum?id=5Q1gr80AXU

  1. Revisiting B2T: Discovering and Mitigating Visual Biases through Keyword Explanations

Faissal El Kayouhi, Aïda Asma, Joey Laarhoven, Fiona Nagelhout

https://openreview.net/forum?id=5GS1q65pv6

  1. Revisiting RegressionMSR: A Reproducibility Study

Neev Mistry

https://openreview.net/forum?id=AzOk6sdMRZ

  1. Stable Continuous-Time Consistency Distillation: An Empirical Study with a Multistep Extension

Vinay Saji Mathew, Soundar Kumara, Gretta Kellogg, William KM Lai

https://openreview.net/forum?id=di6ofoWEU8

  1. WAREX: Web Agent Reliability Evaluation on Existing Benchmarks

Su Kara, Fazle Faisal, Suman Nath

https://openreview.net/forum?id=o4pXVP8RCD

  1. XFeat Revisited: Reproducibility and Evaluation of a Lightweight Image Matcher

Lazar Đoković, Aimee Lin

https://openreview.net/forum?id=2WI889Ulin

We congratulate all the authors, and thank the TMLR action editors and reviewers for their work on these papers. Invited talks for the MLRC session will be announced separately.

We look forward to seeing you in Sydney. Please visit the MLRC 2026 website (https://reproml.org) for full details, and reach out to us at reproducibility-chairs@neurips.cc with any questions.

The MLRC 2026 Organizing Team

Related Posts