Nico Daheim

Sebastian Gehrmann, Abhik Bhattacharjee, Abinaya Mahendiran et al. · amazon-science, cmu

Evaluation in machine learning is usually informed by past choices, for example which datasets or metrics to use. This standardization enables the comparison on equal footing using leaderboards, but the evaluation choices become sub-optimal as better alternatives arise. This problem is especially pertinent in natural language generation which requires ever-improving suites of datasets, metrics, and human evaluation to make definitive claims. To make following best model evaluation practices easier, we introduce GEMv2. The new version of the Generation, Evaluation, and Metrics Benchmark introduces a modular infrastructure for dataset, model, and metric developers to benefit from each others work. GEMv2 supports 40 documented datasets in 51 languages. Models for all datasets can be evaluated online and our interactive data card creation and rendering tools make it easier to add new datasets to the living benchmark.

Jakub Macina, Nico Daheim, Lingzhi Wang et al. · cmu, eth-zurich

Designing dialog tutors has been challenging as it involves modeling the diverse and complex pedagogical strategies employed by human tutors. Although there have been significant recent advances in neural conversational systems using large language models (LLMs) and growth in available dialog corpora, dialog tutoring has largely remained unaffected by these advances. In this paper, we rigorously analyze various generative language models on two dialog tutoring datasets for language learning using automatic and human evaluations to understand the new opportunities brought by these advances as well as the challenges we must overcome to build models that would be usable in real educational settings. We find that although current approaches can model tutoring in constrained learning scenarios when the number of concepts to be taught and possible teacher strategies are small, they perform poorly in less constrained scenarios. Our human quality evaluation shows that both models and ground-truth annotations exhibit low performance in terms of equitable tutoring, which measures learning opportunities for students and how engaging the dialog is. To understand the behavior of our models in a real tutoring setting, we conduct a user study using expert annotators and find a significantly large number of model reasoning errors in 45% of conversations. Finally, we connect our findings to outline future work.

Nico Daheim, Thomas Möllenhoff, Edoardo Maria Ponti et al.

Models trained on different datasets can be merged by a weighted-averaging of their parameters, but why does it work and when can it fail? Here, we connect the inaccuracy of weighted-averaging to mismatches in the gradients and propose a new uncertainty-based scheme to improve the performance by reducing the mismatch. The connection also reveals implicit assumptions in other schemes such as averaging, task arithmetic, and Fisher-weighted averaging. Our new method gives consistent improvements for large language models and vision transformers, both in terms of performance and robustness to hyperparameters. Code available here.

Vilém Zouhar, Shehzaad Dhuliawala, Wangchunshu Zhou et al. · eth-zurich, microsoft-research

Machine translation quality estimation (QE) predicts human judgements of a translation hypothesis without seeing the reference. State-of-the-art QE systems based on pretrained language models have been achieving remarkable correlations with human judgements yet they are computationally heavy and require human annotations, which are slow and expensive to create. To address these limitations, we define the problem of metric estimation (ME) where one predicts the automated metric scores also without the reference. We show that even without access to the reference, our model can estimate automated metrics ($ρ$=60% for BLEU, $ρ$=51% for other metrics) at the sentence-level. Because automated metrics correlate with human judgements, we can leverage the ME task for pre-training a QE model. For the QE task, we find that pre-training on TER is better ($ρ$=23%) than training for scratch ($ρ$=20%).

Nils Dycke, Marina Sakharova, Nico Daheim et al.

Although it is generally agreed that AI-generated text poses a broad societal risk, there is no common understanding in the AI-generated text detection literature on what constitutes harmful use. Rather, existing datasets and approaches often define their own criteria and make their own assumptions, sometimes implicitly, and often only loosely related to real-world needs and applications. To address this gap, we here systematically define various notions of AI-generated text and their characteristics. To study these, we collect AITDNA - a new benchmark of human-machine co-constructed texts that is annotated with detailed genesis information, such as the entire edit and AI-interaction history. We benchmark various machine-generated text detectors and find that they often only perform well for specific notions but not as broad detectors. We release code and data publicly.

Nico Daheim, Jakub Macina, Manu Kapur et al. · eth-zurich

Large language models (LLMs) present an opportunity to scale high-quality personalized education to all. A promising approach towards this means is to build dialog tutoring models that scaffold students' problem-solving. However, even though existing LLMs perform well in solving reasoning questions, they struggle to precisely detect student's errors and tailor their feedback to these errors. Inspired by real-world teaching practice where teachers identify student errors and customize their response based on them, we focus on verifying student solutions and show how grounding to such verification improves the overall quality of tutor response generation. We collect a dataset of 1K stepwise math reasoning chains with the first error step annotated by teachers. We show empirically that finding the mistake in a student solution is challenging for current models. We propose and evaluate several verifiers for detecting these errors. Using both automatic and human evaluation we show that the student solution verifiers steer the generation model towards highly targeted responses to student errors which are more often correct with less hallucinations compared to existing baselines.

Nico Daheim, Nouha Dziri, Mrinmaya Sachan et al.

Ideally, dialogue systems should generate responses that are faithful to the knowledge contained in relevant documents. However, many models generate hallucinated responses instead that contradict it or contain unverifiable information. To mitigate such undesirable behaviour, it has been proposed to fine-tune a `negative expert' on negative examples and subtract its parameters from those of a pre-trained model. However, intuitively, this does not take into account that some parameters are more responsible than others in causing hallucinations. Thus, we propose to weigh their individual importance via (an approximation of) the Fisher Information matrix, which measures the uncertainty of their estimate. We call this method Elastic Weight Removal (EWR). We evaluate our method -- using different variants of Flan-T5 as a backbone language model -- on multiple datasets for information-seeking dialogue generation and compare our method with state-of-the-art techniques for faithfulness, such as CTRL, Quark, DExperts, and Noisy Channel reranking. Extensive automatic and human evaluation shows that EWR systematically increases faithfulness at minor costs in terms of other metrics. However, we notice that only discouraging hallucinations may increase extractiveness, i.e. shallow copy-pasting of document spans, which can be undesirable. Hence, as a second main contribution, we show that our method can be extended to simultaneously discourage hallucinations and extractive responses. We publicly release the code for reproducing EWR and all baselines.

Joris Baan, Nico Daheim, Evgenia Ilia et al.

Recent advances of powerful Language Models have allowed Natural Language Generation (NLG) to emerge as an important technology that can not only perform traditional tasks like summarisation or translation, but also serve as a natural language interface to a variety of applications. As such, it is crucial that NLG systems are trustworthy and reliable, for example by indicating when they are likely to be wrong; and supporting multiple views, backgrounds and writing styles -- reflecting diverse human sub-populations. In this paper, we argue that a principled treatment of uncertainty can assist in creating systems and evaluation protocols better aligned with these goals. We first present the fundamental theory, frameworks and vocabulary required to represent uncertainty. We then characterise the main sources of uncertainty in NLG from a linguistic perspective, and propose a two-dimensional taxonomy that is more informative and faithful than the popular aleatoric/epistemic dichotomy. Finally, we move from theory to applications and highlight exciting research directions that exploit uncertainty to power decoding, controllable generation, self-assessment, selective answering, active learning and more.

David Thulke, Nico Daheim, Christian Dugast et al.

This paper summarizes our contributions to the document-grounded dialog tasks at the 9th and 10th Dialog System Technology Challenges (DSTC9 and DSTC10). In both iterations the task consists of three subtasks: first detect whether the current turn is knowledge seeking, second select a relevant knowledge document, and third generate a response grounded on the selected document. For DSTC9 we proposed different approaches to make the selection task more efficient. The best method, Hierarchical Selection, actually improves the results compared to the original baseline and gives a speedup of 24x. In the DSTC10 iteration of the task, the challenge was to adapt systems trained on written dialogs to perform well on noisy automatic speech recognition transcripts. Therefore, we proposed data augmentation techniques to increase the robustness of the models as well as methods to adapt the style of generated responses to fit well into the proceeding dialog. Additionally, we proposed a noisy channel model that allows for increasing the factuality of the generated responses. In addition to summarizing our previous contributions, in this work, we also report on a few small improvements and reconsider the automatic evaluation metrics for the generation task which have shown a low correlation to human judgments.

Nico Daheim, David Thulke, Christian Dugast et al.

In this work, we present a model for document-grounded response generation in dialog that is decomposed into two components according to Bayes theorem. One component is a traditional ungrounded response generation model and the other component models the reconstruction of the grounding document based on the dialog context and generated response. We propose different approximate decoding schemes and evaluate our approach on multiple open-domain and task-oriented document-grounded dialog datasets. Our experiments show that the model is more factual in terms of automatic factuality metrics than the baseline model. Furthermore, we outline how introducing scaling factors between the components allows for controlling the tradeoff between factuality and fluency in the model output. Finally, we compare our approach to a recently proposed method to control factuality in grounded dialog, CTRL (arXiv:2107.06963), and show that both approaches can be combined to achieve additional improvements.

Junling Wang, Jakub Macina, Nico Daheim et al. · eth-zurich

Educational chatbots are a promising tool for assisting student learning. However, the development of effective chatbots in education has been challenging, as high-quality data is seldom available in this domain. In this paper, we propose a framework for generating synthetic teacher-student interactions grounded in a set of textbooks. Our approaches capture one aspect of learning interactions where curious students with partial knowledge interactively ask a teacher questions about the material in the textbook. We highlight various quality criteria that such dialogues should fulfill and compare several approaches relying on either prompting or fine-tuning large language models. We use synthetic dialogues to train educational chatbots and show benefits of further fine-tuning in different educational domains. However, human evaluation shows that our best data synthesis method still suffers from hallucinations and tends to reiterate information from previous conversations. Our findings offer insights for future efforts in synthesizing conversational data that strikes a balance between size and quality. We will open-source our data and code.

Jakub Macina, Nico Daheim, Ido Hakimi et al. · eth-zurich

Evaluating the pedagogical capabilities of AI-based tutoring models is critical for making guided progress in the field. Yet, we lack a reliable, easy-to-use, and simple-to-run evaluation that reflects the pedagogical abilities of models. To fill this gap, we present MathTutorBench, an open-source benchmark for holistic tutoring model evaluation. MathTutorBench contains a collection of datasets and metrics that broadly cover tutor abilities as defined by learning sciences research in dialog-based teaching. To score the pedagogical quality of open-ended teacher responses, we train a reward model and show it can discriminate expert from novice teacher responses with high accuracy. We evaluate a wide set of closed- and open-weight models on MathTutorBench and find that subject expertise, indicated by solving ability, does not immediately translate to good teaching. Rather, pedagogy and subject expertise appear to form a trade-off that is navigated by the degree of tutoring specialization of the model. Furthermore, tutoring appears to become more challenging in longer dialogs, where simpler questioning strategies begin to fail. We release the benchmark, code, and leaderboard openly to enable rapid benchmarking of future models.

Nico Daheim, Thomas Möllenhoff, Ming Liang Ang et al.

Stochastic Variance Reduced Gradient (SVRG) and its variants aim to speed-up training by using gradient corrections, but have seen limited success in deep learning. Here, we show surprising new foundational connections of SVRG to a recently proposed Bayesian method called posterior correction. Specifically, we show that SVRG is recovered as a special case of posterior correction over the isotropic-Gaussian family, while novel extensions are automatically obtained by using more flexible exponential families. We derive two new SVRG variants by using Gaussian families: First, a Newton-like variant that employs novel Hessian corrections, and second, an Adam-like extension that improves pretraining and finetuning of Transformer language models. This is the first work to connect SVRG to Bayes and use it to boost variational training for deep networks.

Nico Daheim, Clara Meister, Thomas Möllenhoff et al.

Despite their outstanding performance in the majority of scenarios, contemporary language models still occasionally generate undesirable outputs, for example, hallucinated text. While such behaviors have previously been linked to uncertainty, there is a notable lack of methods that actively consider uncertainty during text generation. In this work, we show how Minimum Bayes Risk (MBR) decoding, which selects model generations according to an expected risk, can be generalized into a principled uncertainty-aware decoding method. In short, we account for model uncertainty during decoding by incorporating a posterior over model parameters into MBR's computation of expected risk. We show that this modified expected risk is useful for both choosing outputs and deciding when to abstain from generation and can provide improvements without incurring overhead. We benchmark different methods for learning posteriors and show that performance improves with prediction diversity. We release our code publicly.

Anmol Goel, Nico Daheim, Christian Montag et al.

Reframing a negative into a positive thought is at the crux of several cognitive approaches to mental health and psychotherapy that could be made more accessible by large language model-based solutions. Such reframing is typically non-trivial and requires multiple rationalization steps to uncover the underlying issue of a negative thought and transform it to be more positive. However, this rationalization process is currently neglected by both datasets and models which reframe thoughts in one step. In this work, we address this gap by augmenting open-source datasets for positive text rewriting with synthetically-generated Socratic rationales using a novel framework called \textsc{SocraticReframe}. SocraticReframe uses a sequence of question-answer pairs to rationalize the thought rewriting process. We show that such Socratic rationales significantly improve positive text rewriting for different open-source LLMs according to both automatic and human evaluations guided by criteria from psychotherapy research. We validate our framework and the synthetic rationalizations with expert judgements from domain experts and psychology students in an IRB-approved annotation study. Our findings highlight the potential of utilizing the synergy between LLM reasoning and established psychotherapy techniques to build assistive solutions for reframing negative thoughts.

Bai Cong, Nico Daheim, Yuesong Shen et al.

We show that variational learning can significantly improve the accuracy and calibration of Low-Rank Adaptation (LoRA) without a substantial increase in the cost. We replace AdamW by the Improved Variational Online Newton (IVON) algorithm to finetune large language models. For Llama-2 with 7 billion parameters, IVON improves the accuracy over AdamW by 2.8% and expected calibration error by 4.6%. The accuracy is also better than the other Bayesian alternatives, yet the cost is lower and the implementation is easier. Our work provides additional evidence for the effectiveness of IVON for large language models. The code is available at https://github.com/team-approx-bayes/ivon-lora.

Yuesong Shen, Nico Daheim, Bai Cong et al.

We give extensive empirical evidence against the common belief that variational learning is ineffective for large neural networks. We show that an optimizer called Improved Variational Online Newton (IVON) consistently matches or outperforms Adam for training large networks such as GPT-2 and ResNets from scratch. IVON's computational costs are nearly identical to Adam but its predictive uncertainty is better. We show several new use cases of IVON where we improve finetuning and model merging in Large Language Models, accurately predict generalization error, and faithfully estimate sensitivity to data. We find overwhelming evidence that variational learning is effective.

Artem Shelmanov, Ekaterina Fadeeva, Akim Tsvigun et al.

Large Language Models (LLMs) have the tendency to hallucinate, i.e., to sporadically generate false or fabricated information. This presents a major challenge, as hallucinations often appear highly convincing and users generally lack the tools to detect them. Uncertainty quantification (UQ) provides a framework for assessing the reliability of model outputs, aiding in the identification of potential hallucinations. In this work, we introduce pre-trained UQ heads: supervised auxiliary modules for LLMs that substantially enhance their ability to capture uncertainty compared to unsupervised UQ methods. Their strong performance stems from the powerful Transformer architecture in their design and informative features derived from LLM attention maps. Experimental evaluation shows that these heads are highly robust and achieve state-of-the-art performance in claim-level hallucination detection across both in-domain and out-of-domain prompts. Moreover, these modules demonstrate strong generalization to languages they were not explicitly trained on. We pre-train a collection of UQ heads for popular LLM series, including Mistral, Llama, and Gemma 2. We publicly release both the code and the pre-trained heads.

Falko Helm, Nico Daheim, Iryna Gurevych

Many applications of large language models (LLMs) require long-context understanding, but models continue to struggle with such tasks. We hypothesize that conventional next-token prediction training could contribute to this, because each token is assigned equal weight. Yet, intuitively, the amount of context needed to predict the next token accurately varies greatly across different data. To reflect this, we propose various novel token-weighting schemes that assign different weights to each training token in the loss, thereby generalizing existing works. For this, we categorize token-weighting methods using a two-step framework which compares the confidences of a long-context and short-context model to score tokens. We evaluate all methods on multiple long-context understanding tasks and show that non-uniform loss weights are helpful to improve the long-context abilities of LLMs. Different short-context models can be used effectively for token scoring, including models that are much smaller than the long-context model that is trained. All in all, this work contributes to a better understanding of the trade-offs long-context language modeling faces and provides guidelines for model steering via loss-weighting based on empirical evidence. The code can be found on Github.

Bai Cong, Nico Daheim, Yuesong Shen et al.

Bayesian methods have recently been used to improve LoRA finetuning and, although they improve calibration, their effect on other metrics (such as accuracy) is marginal and can sometimes even be detrimental. Moreover, Bayesian methods also increase computational overheads and require additional tricks for them to work well. Here, we fix these issues by using a recently proposed variational algorithm called IVON. We show that IVON is easy to implement and has similar costs to AdamW, and yet it can also drastically improve many metrics by using a simple posterior pruning technique. We present extensive results on billion-scale LLMs (Llama and Qwen series) going way beyond the scale of existing applications of IVON. For example, we finetune a Llama-3.2-3B model on a set of commonsense reasoning tasks and improve accuracy over AdamW by 1.3% and reduce ECE by 5.4%, outperforming AdamW and other recent Bayesian methods like Laplace-LoRA and BLoB. Overall, our results show that variational learning with IVON can effectively improve LoRA finetuning.

David Dinucu-Jianu, Jakub Macina, Nico Daheim et al. · eth-zurich

Large language models (LLMs) can transform education, but their optimization for direct question-answering often undermines effective pedagogy which requires strategically withholding answers. To mitigate this, we propose an online reinforcement learning (RL)-based alignment framework that can quickly adapt LLMs into effective tutors using simulated student-tutor interactions by emphasizing pedagogical quality and guided problem-solving over simply giving away answers. We use our method to train a 7B parameter tutor model without human annotations which reaches similar performance to larger proprietary models like LearnLM. We introduce a controllable reward weighting to balance pedagogical support and student solving accuracy, allowing us to trace the Pareto frontier between these two objectives. Our models better preserve reasoning capabilities than single-turn SFT baselines and can optionally enhance interpretability through thinking tags that expose the model's instructional planning.

Hugo Monzón Maldonado, Thomas Möllenhoff, Nico Daheim et al.

When finetuning multiple tasks altogether, it is important to carefully weigh them to get a good performance, but searching for good weights can be difficult and costly. Here, we propose to aid the search with fast previews to quickly get a rough idea of different reweighting options. We use model merging to create previews by simply reusing and averaging parameters of models trained on each task separately (no retraining required). To improve the quality of previews, we propose a Bayesian approach to design new merging strategies by using more flexible posteriors. We validate our findings on vision and natural-language transformers. Our work shows the benefits of model merging via Bayes to improve multitask finetuning.

Jakub Macina, Nico Daheim, Sankalan Pal Chowdhury et al.

While automatic dialogue tutors hold great potential in making education personalized and more accessible, research on such systems has been hampered by a lack of sufficiently large and high-quality datasets. Collecting such datasets remains challenging, as recording tutoring sessions raises privacy concerns and crowdsourcing leads to insufficient data quality. To address this, we propose a framework to generate such dialogues by pairing human teachers with a Large Language Model (LLM) prompted to represent common student errors. We describe how we use this framework to collect MathDial, a dataset of 3k one-to-one teacher-student tutoring dialogues grounded in multi-step math reasoning problems. While models like GPT-3 are good problem solvers, they fail at tutoring because they generate factually incorrect feedback or are prone to revealing solutions to students too early. To overcome this, we let teachers provide learning opportunities to students by guiding them using various scaffolding questions according to a taxonomy of teacher moves. We demonstrate MathDial and its extensive annotations can be used to finetune models to be more effective tutors (and not just solvers). We confirm this by automatic and human evaluation, notably in an interactive setting that measures the trade-off between student solving success and telling solutions. The dataset is released publicly.

David Thulke, Nico Daheim, Christian Dugast et al.

This paper summarizes our submission to Task 2 of the second track of the 10th Dialog System Technology Challenge (DSTC10) "Knowledge-grounded Task-oriented Dialogue Modeling on Spoken Conversations". Similar to the previous year's iteration, the task consists of three subtasks: detecting whether a turn is knowledge seeking, selecting the relevant knowledge document and finally generating a grounded response. This year, the focus lies on adapting the system to noisy ASR transcripts. We explore different approaches to make the models more robust to this type of input and to adapt the generated responses to the style of spoken conversations. For the latter, we get the best results with a noisy channel model that additionally reduces the number of short and generic responses. Our best system achieved the 1st rank in the automatic and the 3rd rank in the human evaluation of the challenge.

Nico Daheim, David Thulke, Christian Dugast et al.

This paper summarizes our entries to both subtasks of the first DialDoc shared task which focuses on the agent response prediction task in goal-oriented document-grounded dialogs. The task is split into two subtasks: predicting a span in a document that grounds an agent turn and generating an agent response based on a dialog and grounding document. In the first subtask, we restrict the set of valid spans to the ones defined in the dataset, use a biaffine classifier to model spans, and finally use an ensemble of different models. For the second subtask, we use a cascaded model which grounds the response prediction on the predicted span instead of the full document. With these approaches, we obtain significant improvements in both subtasks compared to the baseline.

David Thulke, Nico Daheim, Christian Dugast et al.

This paper summarizes our work on the first track of the ninth Dialog System Technology Challenge (DSTC 9), "Beyond Domain APIs: Task-oriented Conversational Modeling with Unstructured Knowledge Access". The goal of the task is to generate responses to user turns in a task-oriented dialog that require knowledge from unstructured documents. The task is divided into three subtasks: detection, selection and generation. In order to be compute efficient, we formulate the selection problem in terms of hierarchical classification steps. We achieve our best results with this model. Alternatively, we employ siamese sequence embedding models, referred to as Dense Knowledge Retrieval, to retrieve relevant documents. This method further reduces the computation time by a factor of more than 100x at the cost of degradation in R@1 of 5-6% compared to the first model. Then for either approach, we use Retrieval Augmented Generation to generate responses based on multiple selected snippets and we show how the method can be used to fine-tune trained embeddings.