IFLLM dataset uses mouse and eye-tracking signals to improve LLM alignment via implicit feedback
Researchers introduce IFLLM, a dataset of 1,336 multi-turn interactions from 59 Mechanical Turk workers capturing mouse trajectories and webcam-derived eye gaze to study implicit user feedback for LLM alignment. A reward model trained on this implicit feedback improves text-based reward model accuracy from 55% to 64% and nearly triples relative response quality improvements when combined with DPO across eight LLMs. The work addresses the scarcity and cost of explicit preference annotations by mining behavioral signals already present in user interactions.
Related guides (3)
Related events (8)
RL-based alignment improves interactivity in full-duplex spoken dialogue models
Researchers propose a post-training alignment method using reinforcement learning to improve interactivity in full-duplex spoken dialogue models, which can listen and speak simultaneously. The method addresses four canonical axes of interactivity—pause handling, turn-taking, backchanneling, and user interruption—each with axis-specific reward functions, plus an LLM-based reward to prevent semantic degradation. The approach is applied to two open-source models, Moshi and PersonaPlex, showing consistent improvements in both offline and real-time multi-turn evaluation.
Mitigating Perceptual Judgment Bias in Multimodal LLM-as-a-Judge via Perceptual Perturbation and Reward Modeling
This paper identifies and analyzes 'Perceptual Judgment Bias' in multimodal LLM judges, where models anchor on response text rather than visual evidence when the two conflict. The authors introduce a Perceptually Perturbed Judgment Dataset using counterfactual responses to isolate perceptual errors, and a training framework combining GRPO-based reward modeling with batch-ranking objectives. Experiments on MLLM-as-a-Judge benchmarks show improved perceptual fidelity, ranking coherence, and alignment with human evaluation.
VLMs May Not Globally Enhance Human Alignment over LLMs During Natural Reading
This paper compares matched LLM and VLM pairs in a text-only setting to isolate the effect of multimodal training history on human-like language processing. Using whole-cortex fMRI and eye-tracking data from natural reading, the authors find that multimodal pretraining does not confer a uniform global advantage in human alignment. However, VLMs show selective advantages when sentences contain stronger visual semantic content, with converging evidence from both neural and behavioral measures. The findings suggest language-internal representations remain the primary driver of human text processing alignment.
LLUMI: Fine-Tuning Open-Source LLMs for Mental Health Writing Assistance Using Reddit Community Feedback
LLUMI is a two-component system (a generation model and an improvement model) designed to provide mental health writing assistance using smaller open-source LLMs hosted in privacy-preserving, on-premise environments. The system leverages Reddit community endorsement signals (upvotes/downvotes) to construct preference pairs for SFT and DPO training, then further aligns outputs via human evaluation across readability, empathy, connection, actionability, and safety dimensions. Results show LLUMI achieves performance comparable to proprietary GPT-based models on linguistic and human evaluations, suggesting community-derived preference signals can substitute for expensive expert labeling in sensitive domains.
ExpRL: RL-based mid-training using human QA data as reward scaffolds for LLM reasoning
ExpRL proposes an automated approach to LLM mid-training that replaces manually curated reasoning traces with large corpora of human-written QA data used as reward scaffolds rather than imitation targets. Reference solutions are hidden from the policy and used only to construct problem-specific grading rubrics, enabling dense process-level rewards that reinforce partial progress and intermediate reasoning steps. On challenging math reasoning benchmarks, ExpRL outperforms SFT, sparse-reward GRPO, and self-distillation as an RL initialization strategy, with additional mixed-domain experiments suggesting broader applicability.
FOXGLOVE dataset enables systematic comparison of LLM vs. expert writing feedback on argumentative essays
Researchers introduce FOXGLOVE, a dataset of 2,340 feedback comments on 69 twelfth-grade argumentative essays, comprising 696 comments from trained writing instructors and 1,644 from four frontier LLMs under a shared protocol. The study finds that while instructors and LLMs distribute feedback similarly across goals and essay positions, they diverge on which specific sentences to address. LLM feedback receives higher quality ratings from instructors on most dimensions, but the advantage appears largely attributable to comment length rather than substantive quality. The dataset enables systematic evaluation of human-LLM alignment in educational feedback generation.
Vision Language Model Alignment in TRL
Hugging Face's TRL library has added support for aligning Vision Language Models (VLMs), extending existing RLHF and preference optimization tooling to multimodal settings. The blog post covers the new capabilities for training VLMs with alignment techniques such as DPO and related methods. This expands the open-source ecosystem for multimodal model fine-tuning and alignment.
WhoSaidIt: Human-LLM Collaborative Annotation for Multilingual Speaker-Attribute Classification
This paper proposes a human-LLM collaborative re-annotation framework for stabilizing noisy multilingual speaker-attribute labels under resource constraints. LLMs surface recurring annotation rationales through iterative expert interaction, combined with disagreement-focused sampling for targeted re-annotation. The resulting WhoSaidIt dataset covers nine speaker-attribute labels across multiple languages. Benchmarking of recent LLMs reveals substantial cross-lingual annotation divergence and highlights both capabilities and limitations of LLMs in this classification task.