Title: `GroundCocoa: A Benchmark for Evaluating Compositional & Conditional Reasoning in Language Models`
Abstract: https://arxiv.org/abs/2404.04237
The rapid progress of large language models (LLMs) has seen them excel and frequently surpass human performance on standard benchmarks. This has enabled many downstream applications, such as LLM agents, to rely on their reasoning to address complex task requirements. However, LLMs are known to unexpectedly falter in simple tasks and under seemingly straightforward circumstances - underscoring the need for better and more diverse evaluation setups to measure their true capabilities. To this end, we choose to study compositional and conditional reasoning, two aspects that are central to human cognition, and introduce GroundCocoa - a lexically diverse benchmark connecting these reasoning skills to the real-world problem of flight booking. Our task involves aligning detailed user preferences with available flight options presented in a multiple-choice format. Results indicate a significant disparity in performance among current state-of-the-art LLMs with even the best performing model, GPT-4 Turbo, not exceeding 67% accuracy despite advanced prompting techniques.
title={GroundCocoa: A Benchmark for Evaluating Compositional & Conditional Reasoning in Language Models},
author={Harsh Kohli and Sachin Kumar and Huan Sun},
year={2025},
eprint={2404.04237},
archivePrefix={arXiv},
primaryClass={cs.CL},
url={https://arxiv.org/abs/2404.04237},
}
```
### Groups and Tasks
#### Groups
- Not part of a group yet
#### Tasks
-`groundcocoa`
### Checklist
For adding novel benchmarks/datasets to the library:
* [x] Is the task an existing benchmark in the literature?
* [x] Have you referenced the original paper that introduced the task?
* [x] If yes, does the original paper provide a reference implementation? If so, have you checked against the reference implementation and documented how to run such a test?
If other tasks on this dataset are already supported:
* [ ] Is the "Main" variant of this task clearly denoted?
* [ ] Have you provided a short sentence in a README on what each new variant adds / evaluates?
* [ ] Have you noted which, if any, published evaluation setups are matched by this variant?
question="A user has specified certain criteria for booking a flight. Below are five different flight options labeled 'A', 'B', 'C', 'D', and 'E'. Review these options and select the one that best matches the user requirements. Respond with a single option and the phrase 'The answer is Option ' followed by the correct letter - 'A', 'B', 'C', 'D', or 'E'\n\n"
Title: `Histoires Morales: A French Dataset for Assessing Moral Alignment`
Abstract: `https://arxiv.org/pdf/2501.17117`
⚖ Histoires Morales is the first dataset for moral model alignment evaluation in French. It consists of narratives describing normative and norm-divergent actions taken by individuals to achieve certain intentions in concrete situations, along with their respective consequences.
Each of the 12,000 stories (histoires) follows the same seven-sentence structure as the Moral Stories dataset:
Context:
1. Norm: A guideline for social conduct generally observed by most people in everyday situations.
2. Situation: The setting of the story, introducing participants and describing their environment.
3. Intention: A reasonable goal that one of the story participants (the actor) wants to achieve.
Normative path:
4. Normative action: An action by the actor that fulfills the intention while observing the norm.
5. Normative consequence: A possible effect of the normative action on the actor’s environment.
Norm-divergent path:
6. Divergent action: An action by the actor that fulfills the intention but diverges from the norm.
7. Divergent consequence: A possible effect of the divergent action on the actor’s environment.
Histoires Morales is adapted to French from the widely used Moral Stories dataset.
We translated the Moral Stories dataset and refined these translations through manual annotations.
For adding novel benchmarks/datasets to the library:
* [x] Is the task an existing benchmark in the literature?
* [ ] Have you referenced the original paper that introduced the task?
* [ ] If yes, does the original paper provide a reference implementation? If so, have you checked against the reference implementation and documented how to run such a test?
If other tasks on this dataset are already supported:
* [ ] Is the "Main" variant of this task clearly denoted?
* [ ] Have you provided a short sentence in a README on what each new variant adds / evaluates?
* [ ] Have you noted which, if any, published evaluation setups are matched by this variant?
