Papers by Steven Li

9 papers
CONTOR: Benchmarking Strategies for Completing Ontologies with Plausible Missing Rules (2024.findings-emnlp)

Copied to clipboard

Challenge: Existing evaluations focus on distinguishing held-out ontologies from randomly corrupted ones, which often makes the task unrealistically easy.
Approach: They propose to use the common description logic syntax for encoding ontology rules to test their effectiveness on manually annotated hard negatives.
Outcome: The proposed models are compared with existing models and have been evaluated on different ontologies.
Audio MultiChallenge: A Multi-Turn Evaluation of Spoken Dialogue Systems on Natural Human Interaction (2026.acl-long)

Copied to clipboard

Challenge: End-to-end (E2E) spoken dialogue systems are replacing cascaded pipelines for voice-based human-AI interaction. Existing benchmarks evaluate these systems on synthetic speech and single-turn tasks, leaving multi-turn conversational ability underexplored.
Approach: They propose an open-source benchmark to evaluate spoken dialogue systems under natural multi-turn interaction patterns.
Outcome: The proposed model fails on the highest-performing model with 54.65% pass rate.
What Are We Measuring When We Evaluate Large Vision-Language Models? An Analysis of Latent Factors and Biases (2024.naacl-long)

Copied to clipboard

Challenge: Vision-language models have broad competence that is difficult to evaluate . current evaluation benchmarks focus on only assessing one or a few capabilities .
Approach: They perform a large-scale transfer learning experiment to discover latent VL skills from data.
Outcome: The results suggest that factor analysis can identify reasonable yet surprising VL skill factors . the results contribute to the design of balanced and broad-coverage vision-language evaluation methods.
Two-Stage Fine-Tuning for Improved Bias and Variance for Large Pretrained Language Models (2023.acl-long)

Copied to clipboard

Challenge: Recent work challenges the bias-variance trade-off . large pretrained models can have large variance and overfit domain-specific data .
Approach: They propose a bias-variance trade-off that implies learning methods need to balance complexity with data size to minimize under-fitting and over-fit.
Outcome: The proposed method achieves strong results on SuperGLUE and clinical information extraction tasks.
Modelling Commonsense Commonalities with Multi-Facet Concept Embeddings (2024.findings-acl)

Copied to clipboard

Challenge: Concept embeddings are a useful and efficient mechanism for injecting commonsense knowledge into downstream tasks.
Approach: They propose to model commonalities in concepts by capturing a more diverse range of commonsense properties.
Outcome: The proposed model captures a more diverse range of commonsense properties and improves ontology completion and ultra-fine entity typing tasks.
CodeT5+: Open Code Large Language Models for Code Understanding and Generation (2023.emnlp-main)

Copied to clipboard

Challenge: Existing code LLMs adopt a specific architecture or rely on a unified encoder-decoder network for downstream tasks, lacking flexibility to operate in the optimal architecture for a particular task.
Approach: They propose to initialize code LLMs with frozen off-the-shelf LLM and explore instruction-tuning to align with natural language instructions.
Outcome: The proposed model outperforms open-source LLMs on 20 code-related benchmarks.
What do Deck Chairs and Sun Hats Have in Common? Uncovering Shared Properties in Large Concept Vocabularies (2023.emnlp-main)

Copied to clipboard

Challenge: Existing work on decontextualised concept embeddings from language models has focused on capturing taxonomic structure in concepts.
Approach: They propose a strategy for identifying what different concepts have in common with others and representing them in terms of their properties.
Outcome: The proposed approach improves the performance of state-of-the-art models for a multi-label classification problem.
HyperRouter: Towards Efficient Training and Inference of Sparse Mixture of Experts (2023.emnlp-main)

Copied to clipboard

Challenge: Recent studies suggest that fixing the routers can achieve competitive performance by alleviating the collapsing problem, where all experts eventually learn similar representations.
Approach: They propose a method that dynamically generates router parameters through a fixed hypernetwork and trainable embeddings to achieve a balance between training the routers and freezing them to learn an improved routing policy.
Outcome: Experiments on a wide range of tasks show that the proposed method performs better than existing methods.
Ultra-Fine Entity Typing with Prior Knowledge about Labels: A Simple Clustering Based Strategy (2023.findings-emnlp)

Copied to clipboard

Challenge: Ultra-fine entity typing is a task of inferring the semantic types from a large set of fine-grained candidates that apply to a given entity mention.
Approach: They propose to use pre-trained label embeddings to cluster the labels into semantic domains and treat them as additional types.
Outcome: The proposed method improves the performance of existing models with high quality embeddings.

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