Entity · paper

Controlling Implicit Shortcut Reliance in L2 Spoken English Auto-markers

paperactiveprovisionalcontrolling-implicit-shortcut-reliance-in-l2-spoken-english-auto-markers-2e67eb1b·1 events·first seen 43h ago

Aliases: Controlling Implicit Shortcut Reliance in L2 Spoken English Auto-markers

More like this (12)

Auditing Protocol-Level Shortcuts in Large Audio Language Model Judges for Speech Evaluation language-adaptive switch Progressive Refinement: An Iterative Pseudo-Labeling Approach for Mandarin-English Code-Switching ASR Prompting Complexity: Shortest Prompts for Texts and Behaviors in LLMs Learning to Hear Hesitation: Continual Learning for Disfluency-Aware ASR Can We Break LLMs Out of Self-Loops? Fine-Grained Reasoning Control with Activation Steering PAT (Pragmatic Auto-Translator)On the Limits of Prompt-Conditioned Language Models as General-Purpose Learners Accuracy and Satisfaction in Multi-Turn LLM Dialogues for NFR Assessment Actionable Activation Directions for Detecting and Mitigating Emergent Misalignment Across Language Model Families Revising Context, Shifting Simulated Stance: Auditing LLM-Based Stance Simulation in Online Discussions How Do Instructions Shape Speech? Cross-Attention Attribution for Style-Captioned Text-to-Speech

Recent events (1)

3arXiv · cs.CL·43h ago·source ↗

Novel training criterion reduces shortcut reliance in L2 spoken English automated scoring systems

A new arXiv paper introduces a training criterion designed to reduce shortcut reliance in automated language proficiency assessment systems, targeting both audio-based and ASR-text-based scoring pipelines. The authors demonstrate that transformer-based auto-markers over-rely on exploitable features relative to human raters, creating malpractice opportunities for test-takers. The proposed criterion reduces this over-reliance, bringing model behavior closer to human reference correlations.

Evaluation and Benchmarking Controlling Implicit Shortcut Reliance in L2 Spoken English Auto-markers