Papers by Flora D. Salim
Why Supervised Fine-Tuning Fails to Learn: A Systematic Study of Incomplete Learning in Large Language Models (2026.acl-long)
Copied to clipboard
Chao Xue, Yao Wang, Mengqiao Liu, Di Liang, Xingsheng Han, Peiyang Liu, Xianjie Wu, Chenyao Lu, Lei Jiang, Yu Lu, Haibo Shi, Shuang Liang, Minlong Peng, Flora D. Salim
| Challenge: | Incomplete learning is widespread and heterogeneous in large language models . authors identify five recurrent sources of incomplete learning: missing prerequisite knowledge, conflicts between SFT supervision and pre-training knowledge, internal inconsistencies within SFT data, left-side forgetting during sequential fine-tuning, and insufficient optimization for rare or complex patterns. |
| Approach: | They propose a diagnostic-first framework that maps incomplete learning to causes . they identify five recurrent sources of incomplete learning: missing prerequisite knowledge, conflicts between supervision and pre-training knowledge, internal inconsistencies, left-side forgetting during sequential fine-tuning, and insufficient optimization for rare or complex patterns. |
| Outcome: | The proposed framework maps incomplete learning to causes using observable training and inference signals. |
Long Context Modeling with Ranked Memory-Augmented Retrieval (2026.acl-long)
Copied to clipboard
| Challenge: | Large Language Models (LLMs) face a fundamental limitation in processing long-context scenarios due to quadratic complexity of attention mechanisms and increasing memory demands during generation. |
| Approach: | They propose a framework that dynamically ranks memory entries based on relevance . ERMAR employs a relevance scoring mechanism and a pointwise re-ranking model for key-value embeddings . |
| Outcome: | The proposed framework achieves state-of-the-art performance on benchmarks . it uses historical usage patterns and adaptive retrieval to improve performance . |
Reason Only When Needed: Efficient Generative Reward Modeling via Model-Internal Uncertainty (2026.findings-acl)
Copied to clipboard
Chao Xue, Yao Wang, Mengqiao Liu, Di Liang, Xingsheng Han, Peiyang Liu, Xianjie Wu, Chenyao Lu, Lei Jiang, Yu Lu, Haibo Shi, Shuang Liang, Minlong Peng, Flora D. Salim
| Challenge: | Existing approaches to generating reward models rely on voting-based mechanisms to evaluate CoT outputs. |
| Approach: | They propose an efficient generative reward modeling framework grounded in model-internal uncertainty. |
| Outcome: | The proposed framework reduces inference cost while improving answer accuracy. |
SensorLLM: Aligning Large Language Models with Motion Sensors for Human Activity Recognition (2025.emnlp-main)
Copied to clipboard
| Challenge: | SensorLLM is a timeseries classification framework that can perform human activity recognition tasks. |
| Approach: | They propose a framework that enables Large Language Models to perform human activity recognition from sensor time-series data. |
| Outcome: | The proposed framework can perform human activity recognition (HAR) tasks with human inputs. |
SOCIA-EVO: Automated Simulator Construction via Dual-Anchored Bi-Level Optimization (2026.acl-long)
Copied to clipboard
| Challenge: | Large Language Models (LLMs) demonstrate strong capabilities in translating natural language into code, but applying them to this domain remains challenging. |
| Approach: | They propose a dual-anchored evolutionary framework that combines a static blueprint and a bi-level optimization to decouple structural refinement from parameter calibration. |
| Outcome: | The proposed framework identifies two failure modes in long-horizon LLM agents: contextual drift and optimization instability arising from conflating structural and parametric errors. |
Beyond Words: Integrating Theory of Mind into Conversational Agents for Human-Like Belief, Desire, and Intention Alignment (2025.findings-acl)
Copied to clipboard
| Challenge: | Empirical evaluations of LLaMA-3 models demonstrate that ToM-informed alignment improves response quality, achieving win rates of 63% and 67%, respectively. |
| Approach: | They investigate whether open-source LLaMA models can represent and retain ToM-related constructs and whether they can be used to generate more aligned responses. |
| Outcome: | The proposed models can represent and retain ToM-related constructs and improve response quality. |
ZARA: Training-Free Motion Time-Series Reasoning via Evidence-Grounded LLM Agents (2026.acl-long)
Copied to clipboard
| Challenge: | Existing approaches to human activity recognition are constrained to fixed activity sets . lack of training-free adaptation to new behavior leads to hallucinations and weak grounding . |
| Approach: | They propose a knowledge- and retrieval-augmented agentic framework for motion time-series reasoning in a training-free inference setting. |
| Outcome: | The proposed framework generalizes robustly to unseen subjects and across datasets . it can be used to train-free inference in a training-free environment . |