Papers by Sharon Li

8 papers
How Retrieved Context Shapes Internal Representations in RAG (2026.findings-acl)

Copied to clipboard

Challenge: Retrieval-augmented generation (RAG) is a widely adopted approach for enhancing large language models with external knowledge.
Approach: They analyze how different types of retrieved documents affect the hidden states of large language models and how these internal representation shifts relate to downstream generation behavior.
Outcome: The results show that context relevancy and layer-wise processing influence internal representations, providing explanations of LLMs’ output behaviors and insights for RAG system design.
ModeX: Evaluator-Free Best-of-N Selection for Open-Ended Generation (2026.acl-long)

Copied to clipboard

Challenge: Existing approaches to selecting a single high-quality output from multiple generations are limited in their applicability and efficiency.
Approach: They propose a method that generalizes majority voting to open-ended text generation . modeX leverages structural information across multiple generation paths to select a "modal" output .
Outcome: The proposed framework outperforms standard single- and multi-path baselines in open-ended tasks.
GeoArena: Evaluating Open-World Geographic Reasoning in Large Vision-Language Models (2026.acl-long)

Copied to clipboard

Challenge: Existing evaluation paradigms for geographic reasoning are outcome-centric and focus on label matching, leaving the underlying linguistic reasoning chains as unexamined black boxes.
Approach: They propose a dynamic, human-preference-based evaluation framework for benchmarking open-world geographic reasoning.
Outcome: The proposed framework reframes evaluation as a pairwise reasoning alignment task on in-the-wild images, where human judges compare model-generated explanations based on reasoning quality, evidence synthesis, and plausibility.
Why Multimodal In-Context Learning Lags Behind? Unveiling the Inner Mechanisms and Bottlenecks (2026.acl-long)

Copied to clipboard

Challenge: In-context learning (ICL) enables models to adapt to new tasks via inference-time demonstrations.
Approach: They propose a simple inference-stage enhancement method that reinforces task mapping transfer.
Outcome: The proposed method strengthens task mapping transfer in multimodal models . it performs comparable to text-only ICL in zero-shot settings but degrades significantly under few-shot demonstrations.
When Identity Skews Debate: Anonymization for Bias-Reduced Multi-Agent Reasoning (2026.acl-long)

Copied to clipboard

Challenge: Multi-agent debate (MAD) aims to improve large language model reasoning by letting multiple agents exchange answers and then aggregate their opinions.
Approach: They propose a principled framework that joins sycophancy and self-bias to mitigate and quantify identity bias in multi-agent debate by removing identity markers from prompts.
Outcome: The proposed framework joins identity-driven sycophancy and self-bias to mitigate and quantify identity bias in multi-agent debate.
LAD: Learning Advantage Distribution for Reasoning (2026.findings-acl)

Copied to clipboard

Challenge: Recent reinforcement learning objectives focus on maximizing expected rewards, limiting diversity and exploration.
Approach: They propose a distribution-matching framework that replaces advantage maximization with learning the advantage-induced distribution.
Outcome: Experiments on math and code reasoning tasks show that LAD improves accuracy and diversity.
Uncertainty Quantification in LLM Agents: Foundations, Emerging Challenges, and Opportunities (2026.acl-long)

Copied to clipboard

Challenge: Uncertainty quantification (UQ) for large language models is a key building block for daily applications.
Approach: They propose a general formulation of agent UQ that subsumes broad classes of existing UQ setups.
Outcome: The proposed framework is based on the first general formulation of agent UQ that subsumes broad classes of existing setups.
VAUQ: Vision-Aware Uncertainty Quantification for LVLM Self-Evaluation (2026.findings-acl)

Copied to clipboard

Challenge: Existing self-evaluation methods rely on a model’s ability to estimate the correctness of its own outputs, but they depend heavily on language priors and are therefore ill-suited for evaluating vision-conditioned predictions.
Approach: They propose a vision-aware uncertainty quantification framework that measures how strongly a model’s output depends on visual evidence.
Outcome: The proposed framework outperforms existing methods across multiple datasets.

What is GenGO?

GenGO is an NLP powered publication search system. It currenctly indexes 30k+ papers from ACL Anthology, and implements multi-aspect summarization, semantic search, and more!

Information

About
Limitations