Entity · paper

Groc-PO: Grounded Context Preference Optimization for Truthful Multimodal LLMs

paperactiveprovisionalgroc-po-grounded-context-preference-optimization-for-truthful-multimodal-llms-d1bb9fe6·1 events·first seen 4d ago

Aliases: Groc-PO: Grounded Context Preference Optimization for Truthful Multimodal LLMs

Co-occurring entities

Grounded Context Preference Dataset Direct Preference Optimization (DPO)Groc-PO

More like this (12)

Training LLMs to Enforce Multi-Level Instruction Hierarchies via Gravity-Weighted Direct Preference Optimization From Correctness to Utility: Gain-Based Prefix Evaluation for LLM Reasoning GraphPO: Graph-based Policy Optimization for Reasoning Models ProvenanceGuard: Source-Aware Factuality Verification for MCP-Based LLM Agents Knowledge Knows, Verbalization Tells: Disentangling Latent Directions for Mathematical Solvability in LLMs long-context LLMs Operadic consistency: a label-free signal for compositional reasoning failures in LLMs Prompting Complexity: Shortest Prompts for Texts and Behaviors in LLMs Janus: A Benchmark for Goal-Conditioned Information Distortion in LLMs Reinforcement Learning with Metacognitive Feedback Elicits Faithful Uncertainty Expression in LLMs Can LLMs Judge Better Than They Generate? Evaluating Task Asymmetry, Mechanistic Interpretability and Transferability for In-Context QA Conversable Complexity: Agentic LLM Collectives as Interpretable Substrates

Recent events (1)

5arXiv · cs.CL·4d ago·source ↗

Groc-PO: Grounded Context Preference Optimization reduces hallucination in multimodal LLMs

Researchers introduce Groc-PO, a preference optimization framework for multimodal LLMs that applies stage-specific supervision across object grounding, contextual grounding, and grounded reasoning stages, rather than only at the final-answer level as in standard DPO. The method is paired with a new dataset (GCPD) organized around these three grounding stages. Experiments show improvements over standard DPO and other baselines on hallucination mitigation and faithful reasoning, addressing the credit-assignment problem where early grounding errors propagate through reasoning chains.

Alignment and RLHF Multimodal Progress Grounded Context Preference Dataset Direct Preference Optimization (DPO)Groc-PO: Grounded Context Preference Optimization for Truthful Multimodal LLMs +1 more