Papers by Mario Fritz
Accept or Deny? Evaluating LLM Fairness and Performance in Loan Approval across Table-to-Text Serialization Approaches (2025.findings-emnlp)
Copied to clipboard
Israel Abebe Azime, Deborah D. Kanubala, Tejumade Afonja, Mario Fritz, Isabel Valera, Dietrich Klakow, Philipp Slusallek
| Challenge: | Large Language Models (LLMs) are increasingly employed in high-stakes decision-making tasks such as loan approvals. |
| Approach: | They evaluate the performance and fairness of LLMs on serialized loan approval datasets from Ghana, Germany, and the United States. |
| Outcome: | The model’s zero-shot and in-context learning (ICL) capabilities are evaluated on loan approval datasets from Ghana, Germany, and the United States. |
Balancing Diversity and Risk in LLM Sampling: How to Select Your Method and Parameter for Open-Ended Text Generation (2025.acl-long)
Copied to clipboard
| Challenge: | Recent studies suggest that sampling-based decoding strategies can be used to optimize the output of Large Language Models (LLMs) . previous studies have shown that likelihood-maximization produces degenerate text which contains repetitive loops and incoherent context, especially in open-ended tasks. |
| Approach: | They propose to use a prefix tree to estimate the intrinsic capacity of a truncation sampling method by considering the trade-off between diversity and risk at each decoding step. |
| Outcome: | The proposed method is based on a prefix tree which preserves the context of a full sentence. |
LLM Task Interference: An Initial Study on the Impact of Task-Switch in Conversational History (2024.emnlp-main)
Copied to clipboard
| Challenge: | Recent advances in Natural Language Processing (NLP) have led to the widespread deployment of large language models (LLMs) across various applications. |
| Approach: | They propose to formalize the study of task-switches in conversational LLMs by analyzing conversational history. |
| Outcome: | The proposed study formalizes and investigates the sensitivity of large language models to taskswitch scenarios in conversational LLMs. |
ProxyPrompt: Securing System Prompts against Prompt Extraction Attacks (2026.findings-acl)
Copied to clipboard
| Challenge: | Existing defenses against prompt extraction are either easily bypassed or require constant updates to address new threats. |
| Approach: | They propose a new mechanism that replaces the original prompt with a proxy to prevent prompt leakage by obfuscating the extracted prompt. |
| Outcome: | The proposed defense outperforms the existing defense, which only achieves 42.80% of the prompts extracted from the original task. |
SimSCOOD: Systematic Analysis of Out-of-Distribution Generalization in Fine-tuned Source Code Models (2024.findings-naacl)
Copied to clipboard
| Challenge: | Large datasets are increasingly available for pre-training source code models, but obtaining representative training data that fully covers the code distribution for specific downstream tasks remains challenging due to the task-specific nature and limited labeling resources. |
| Approach: | They propose a systematic approach that simulates various OOD scenarios along different dimensions of source code data properties and investigates model behavior under different fine-tuning methodologies. |
| Outcome: | The proposed approach simulates various OOD scenarios along different dimensions of source code data properties and exposes multiple failure modes attributed to OOD generalization issues. |
Context-Aware Reasoning On Parametric Knowledge for Inferring Causal Variables (2025.findings-emnlp)
Copied to clipboard
| Challenge: | randomized experiments provide strong inferences, but are often infeasible due to ethical or practical constraints. |
| Approach: | They propose a benchmark where the objective is to complete a partial causal graph. |
| Outcome: | The proposed benchmarks show that they can hypothesize backdoor variables between a cause and its effect. |
PoLLMgraph: Unraveling Hallucinations in Large Language Models via State Transition Dynamics (2024.findings-naacl)
Copied to clipboard
| Challenge: | Existing studies have recognized hallucination as a notable concern in large autoregressive language models (LLMs). |
| Approach: | They propose a polygraph for large language models that detects "hallucination" they demonstrate that hallucination can be detected by tractable probabilistic models . |
| Outcome: | The proposed model outperforms state-of-the-art methods on open-source LLMs by 20% on TruthfulQA benchmarks. |
A Theory of Response Sampling in LLMs: Part Descriptive and Part Prescriptive (2025.acl-long)
Copied to clipboard
| Challenge: | Large Language Models (LLMs) are increasingly utilized in autonomous decision-making, where they sample options from vast action spaces. |
| Approach: | They propose to use heuristics to sample LLMs to identify a prescriptive component and a descriptive component to represent a concept. |
| Outcome: | The proposed model is compared with human decision-making models in public health, economic trends and other real-world domains to show that it is biased. |
CausalGraph2LLM: Evaluating LLMs for Causal Queries (2025.findings-naacl)
Copied to clipboard
| Challenge: | Recent advances in Large Language Models (LLMs) have opened up new avenues for their use beyond standard Natural Language Processing tasks. |
| Approach: | They propose a benchmark to evaluate the capabilities of Large Language Models (LLMs) they use over 700k queries to compare their encoding capabilities. |
| Outcome: | The proposed benchmark compared LLMs on graph-level and node-level queries and open-sourced and closed models. |
Funny or Persuasive, but Not Both: Evaluating Fine-Grained Multi-Concept Control in LLMs (2026.eacl-short)
Copied to clipboard
| Challenge: | Large Language Models (LLMs) provide strong generative capabilities, but many applications require explicit and fine-grained control over specific textual concepts. |
| Approach: | They propose a framework for fine-grained controllability for single- and dual-concept scenarios . they find performance drops in the dual-constituency setting, even though chosen concepts should be separable . |
| Outcome: | The proposed framework shows that models struggle with compositionality even when concepts are intuitively independent. |
Justice in Judgment: Unveiling (Hidden) Bias in LLM-assisted Peer Reviews (2026.findings-acl)
Copied to clipboard
| Challenge: | Existing studies show that large language models carry implicit biases across race, gender, and religion . prior studies documented such biase based on text generation and classification tasks . |
| Approach: | They investigate bias in large language models by controlling metadata on author metadata . authors found affiliation bias favoring authors from highly ranked institutions . |
| Outcome: | The proposed model favors authors from highly ranked institutions, the authors show . the model also favors author affiliations from highly-ranked institutions . |