GRIP (Grounded Reasoning via Information-Restricted Premises), which imposes capacity asymmetry: the decoder keeps full-dimensional access to the query, while retrieved evidence passes through a severe stochastic bottleneck, which forces the evidence channel to encode only the residual information unavailable from the query.
Abstract
High-capacity encoders in retrieval-augmented generation (RAG) can let the query dominate the latent state, leaving retrieved evidence functionally irrelevant. We call this failure mode query dominance. To address it, we introduce \textbf{GRIP} (Grounded Reasoning via Information-Restricted Premises), which imposes capacity asymmetry: the decoder keeps full-dimensional access to the query, while retrieved evidence passes through a severe stochastic bottleneck. This forces the evidence channel to encode only the residual information unavailable from the query. Across five reasoning benchmarks, GRIP outperforms strong iterative baselines, cuts a query--latent mutual-information diagnostic by roughly 30$\times$ (14.8 $\to$ 0.47 bits), and reduces hallucination by 73\%. Residual-alignment analysis further shows that the bottleneck output occupies subspaces less aligned with the query than baseline representations.
GLaQ, a grounded latent-query framework that replaces sequential latent rollout with a fixed set of context-conditioned queries grounded in the original visual tokens, is introduced, suggesting that direct query-to-image grounding can recover localized evidence from the full image without external visual operations or...
Ze-Sheng Yang, Lingling Zhang, Xinyu Zhang et al.· 1 citation
Retrieval-Augmented Generation (RAG) has emerged as the dominant strategy to ground LLM outputs in retrieved evidence to reduce hallucinations and increase factuality, yet RAG itself introduces new challenges: noisy retrieval, knowledge conflicts, and wasted computation from unnecessary retrieval. This tutorial provide...
X. Dong, Sanat Sharma, Kai Sun et al.· Proceedings of the 32nd ACM...· 0 citations
Experiments show that Search-G1 improves the grounding--search-cost trade-off, producing shorter response-side trajectories at competitive task accuracy and periodically refits both readouts on trajectories from the latest checkpoint, allowing the reward to co-evolve with the policy.
Ruoxi Cheng, Haoxuan Ma, Hongyi Zhang et al.· 0 citations
UnicDocVLM is proposed, an end-to-end framework that unifies OCR and visual RAG within a single vision-language model: the model first generates a structured parse of retrieved pages, then activates question-relevant evidence from the parse to support grounded reasoning and answering.
Zongsheng Cao, Anran Liu, Jun Xie et al.· Proceedings of the 32nd ACM...· 0 citations
Consistent gains over DVD on LVBench, Video-MME, EgoSchema, and LongVideoBench suggest that option-aware evidence acquisition transfers beyond MMR-V, and proposes PACE (Progressive Acquisition of Critical Evidence), a factor-guided framework for long-video evidence acquisition.
Bai-Xuan Xu, Yinyui Xu, Tianshi ZHENG et al.· 1 citation
Despite-encoder vision-language models expose a similarity interface that enables zero-shot retrieval but fails compositional constraints, this work proposes factored inference, which separates evidence extraction from constraint execution, and introduces LCSE (Logic-Constrained Score Editing), a training-free method t...
S. Alshehri, Zhan-Tao Yang, Han Zhang et al.· arXiv.org· 0 citations
Machine-learning networks can learn without detailed instructions from a designer. Those that do may offer insight into how the brain reorganizes itself during learning.
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
