Entity · benchmark
MMLU-Pro
benchmarkactive
mmlu-pro-2baddf83·5 events·first seen Aliases: MMLU-Pro
Co-occurring entities
Uncertainty-based Debiasing and Unlearning for DecontaminationMATH-MCQAGemma 2 9Bassistant axisLlama 3.1 70BEQ-BenchDeepSeek V4ML Alignment & Theory Scholars Programactivation cappingQwen3 32BUniversity of OxfordIFEvalChristina LuGSM8KAnthropicCohen's hG*Powerresolution ratio qanytime-valid sequential testingOpen LLM Leaderboard
More like this (12)
Recent events (5)
Uncertainty-Based Decontamination (UBD) framework for removing benchmark contamination from LLMs
Researchers propose Uncertainty-Based Decontamination (UBD), a method that uses deep ensembles of a contaminated model to estimate per-sample memorization and correct for benchmark data contamination without requiring access to an uncontaminated reference model. The approach introduces a sample-level evaluation framework using distributional distance metrics alongside aggregate accuracy to better characterize decontamination quality. Experiments on MMLU-Pro and MATH-MCQA show UBD produces output distributions closer to uncontaminated baselines than paraphrasing or choice-permutation methods. The work addresses a significant validity concern in LLM evaluation, where contamination inflates reported benchmark performance.
Activation Capping Technique Stabilizes LLM Assistant Personas Against Drift and Jailbreaks
Researchers from MATS, Oxford, and Anthropic introduced the 'assistant axis,' a vector derived from LLM layer outputs that quantifies how closely a model adheres to its trained assistant persona. They developed 'activation capping,' an inference-time method that corrects deviations from this axis when similarity falls below a threshold. Testing on Gemma 2 27B, Qwen3 32B, and Llama 3.3 70B showed harmful response rates to jailbreak prompts dropped by roughly half (e.g., 83% to 41% for Qwen3 32B) without degrading benchmark performance. The technique targets character-based jailbreaks that bypass system prompts by manipulating a model's internal representational state.
Resolution Diagnostics for Paired LLM Evaluation: Many Leaderboard Rankings Statistically Unresolved
This paper frames pairwise LLM evaluation as a hypothesis-testing problem and introduces a resolution ratio q = N/N* to diagnose whether leaderboard comparisons are statistically powered. Applying this to two public leaderboards, the authors find that 11/40 Open LLM Leaderboard v1 pairwise comparisons and 4-6/9 MMLU-Pro top-10 adjacent-rank pairs fail to meet conventional (alpha=0.05, power=0.8) resolution targets. A key finding is that the widely-used unpaired Cohen-h shortcut underestimates required sample size by approximately a factor of two in close-comparison regimes, a flaw silently inherited by three major statistical calculators. The unresolved-pair pattern persists under multiplicity correction and sequential testing.
Training-Free Looped Transformers: Inference-Time Recurrence via ODE-Motivated Layer Reapplication
The paper introduces a method to retrofit recurrence onto frozen pretrained transformer checkpoints at inference time by looping a contiguous mid-stack block of layers without any fine-tuning or architectural changes. Naive block reapplication degrades performance, so the authors motivate their approach by treating pre-norm transformer blocks as forward Euler ODE steps and replacing one large update with smaller damped sub-steps. Evaluated across seven model families including dense, sparse MoE, and MLA+MoE architectures, the method yields consistent benchmark improvements (e.g., +2.64 pp on MMLU-Pro for Qwen3-4B-Instruct) at no training cost.
Data Points: Thinking Machines Interaction Model, ERNIE 5.1, Co-Mathematician, RL Conductor, and More
This edition of The Batch covers five notable AI developments: Thinking Machines' research preview of an 'interaction model' with a 200ms micro-turn multimodal architecture; Baidu's ERNIE 5.1, a compressed derivative of ERNIE 5.0 using only 6% of typical pre-training compute; Google DeepMind's Co-Mathematician collaborative workbench reaching 48% on FrontierMath Tier 4; a 7B RL Conductor model that orchestrates multi-agent workflows via reinforcement learning; and Google's Magic Pointer cursor system powered by Gemini. Secondary items include GitHub Copilot pricing restructuring ahead of usage-based billing.