On Initializing Transformers with Pre-trained Embeddings

It has become common practice now to use random initialization schemes, rather than the pre-trained embeddings, when training transformer based models from scratch. Indeed, we find that pre-trained word embeddings from GloVe, and some sub-word embeddings extracted from language models such as T5 and mT5 fare much worse compared to random initialization. This is counter-intuitive given the well-known representational and transfer-learning advantages of pre-training. Interestingly, we also find that BERT and mBERT embeddings fare better than random initialization, showing the advantages of pre-trained representations. In this work, we posit two potential factors that contribute to these mixed results: the model sensitivity to parameter distribution and the embedding interactions with position encodings. We observe that pre-trained GloVe, T5, and mT5 embeddings have a wider distribution of values. As argued in the initialization studies, such large value initializations can lead to poor training because of saturated outputs. Further, the larger embedding values can, in effect, absorb the smaller position encoding values when added together, thus losing position information. Standardizing the pre-trained embeddings to a narrow range (e.g. as prescribed by Xavier) leads to substantial gains for Glove, T5, and mT5 embeddings. On the other hand, BERT pre-trained embeddings, while larger, are still relatively closer to Xavier initialization range which may allow it to effectively transfer the pre-trained knowledge.

CaT-BENCH: Benchmarking Language Model Understanding of Causal and Temporal Dependencies in Plans

Understanding the abilities of LLMs to reason about natural language plans, such as instructional text and recipes, is critical to reliably using them in decision-making systems. A fundamental aspect of plans is the temporal order in which their steps needs to be executed, which reflects the underlying causal dependencies between them. We introduce CaT-Bench, a benchmark of Step Order Prediction questions, which test whether a step must necessarily occur before or after another in cooking recipe plans. We use this to evaluate how well frontier LLMs understand causal and temporal dependencies. We find that SOTA LLMs are underwhelming (best zero-shot is only 0.59 in F1), and are biased towards predicting dependence more often, perhaps relying on temporal order of steps as a heuristic. While prompting for explanations and using few-shot examples improve performance, the best F1 result is only 0.73. Further, human evaluation of explanations along with answer correctness show that, on average, humans do not agree with model reasoning. Surprisingly, we also find that explaining after answering leads to better performance than normal chain-of-thought prompting, and LLM answers are not consistent across questions about the same step pairs. Overall, results show that LLMs' ability to detect dependence between steps has significant room for improvement.

Comparing Human-Centered Language Modeling: Is it Better to Model Groups, Individual Traits, or Both?

Natural language processing has made progress in incorporating human context into its models, but whether it is more effective to use group-wise attributes (e.g., over-45-year-olds) or model individuals remains open. Group attributes are technically easier but coarse: not all 45-year-olds write the same way. In contrast, modeling individuals captures the complexity of each person's identity. It allows for a more personalized representation, but we may have to model an infinite number of users and require data that may be impossible to get. We compare modeling human context via group attributes, individual users, and combined approaches. Combining group and individual features significantly benefits user-level regression tasks like age estimation or personality assessment from a user's documents. Modeling individual users significantly improves the performance of single document-level classification tasks like stance and topic detection. We also find that individual-user modeling does well even without user's historical data.

Modeling Complex Event Scenarios via Simple Entity-focused Questions

Event scenarios are often complex and involve multiple event sequences connected through different entity participants. Exploring such complex scenarios requires an ability to branch through different sequences, something that is difficult to achieve with standard event language modeling. To address this, we propose a question-guided generation framework that models events in complex scenarios as answers to questions about participants. At any step in the generation process, the framework uses the previously generated events as context, but generates the next event as an answer to one of three questions: what else a participant did, what else happened to a participant, or what else happened. The participants and the questions themselves can be sampled or be provided as input from a user, allowing for controllable exploration. Our empirical evaluation shows that this question-guided generation provides better coverage of participants, diverse events within a domain, comparable perplexities for modeling event sequences, and more effective control for interactive schema generation.

Interleaving Retrieval with Chain-of-Thought Reasoning for Knowledge-Intensive Multi-Step Questions

Recent work has shown that large language models are capable of generating natural language reasoning steps or Chains-of-Thoughts (CoT) to answer a multi-step question when prompted to do so. This is insufficient, however, when the necessary knowledge is not available or up-to-date within a model's parameters. A straightforward approach to address this is to retrieve text from an external knowledge source using the question as a query and prepend it as context to the model's input. This, however, is also insufficient for multi-step QA where \textit{what to retrieve} depends on \textit{what has already been derived}. To address this issue we propose IRCoT, a new approach that interleaves retrieval with CoT for multi-step QA, guiding the retrieval with CoT and in turn using retrieved results to improve CoT. Our experiments with GPT3 show substantial improvements in retrieval (up to 22 points) and downstream QA (up to 16 points) over the baselines on four datasets: HotpotQA, 2WikiMultihopQA, MuSiQue, and IIRC. Notably, our method also works well for much smaller models such as T5-Flan-large (0.7B) without any additional training.

POQue: Asking Participant-specific Outcome Questions for a Deeper Understanding of Complex Events

Knowledge about outcomes is critical for complex event understanding but is hard to acquire. We show that by pre-identifying a participant in a complex event, crowd workers are able to (1) infer the collective impact of salient events that make up the situation, (2) annotate the volitional engagement of participants in causing the situation, and (3) ground the outcome of the situation in state changes of the participants. By creating a multi-step interface and a careful quality control strategy, we collect a high quality annotated dataset of 8K short newswire narratives and ROCStories with high inter-annotator agreement (0.74-0.96 weighted Fleiss Kappa). Our dataset, POQue (Participant Outcome Questions), enables the exploration and development of models that address multiple aspects of semantic understanding. Experimentally, we show that current language models lag behind human performance in subtle ways through our task formulations that target abstract and specific comprehension of a complex event, its outcome, and a participant's influence over the event culmination.

BioNLI: Generating a Biomedical NLI Dataset Using Lexico-semantic Constraints for Adversarial Examples

Natural language inference (NLI) is critical for complex decision-making in biomedical domain. One key question, for example, is whether a given biomedical mechanism is supported by experimental evidence. This can be seen as an NLI problem but there are no directly usable datasets to address this. The main challenge is that manually creating informative negative examples for this task is difficult and expensive. We introduce a novel semi-supervised procedure that bootstraps an NLI dataset from existing biomedical dataset that pairs mechanisms with experimental evidence in abstracts. We generate a range of negative examples using nine strategies that manipulate the structure of the underlying mechanisms both with rules, e.g., flip the roles of the entities in the interaction, and, more importantly, as perturbations via logical constraints in a neuro-logical decoding system. We use this procedure to create a novel dataset for NLI in the biomedical domain, called BioNLI and benchmark two state-of-the-art biomedical classifiers. The best result we obtain is around mid 70s in F1, suggesting the difficulty of the task. Critically, the performance on the different classes of negative examples varies widely, from 97% F1 on the simple role change negative examples, to barely better than chance on the negative examples generated using neuro-logic decoding.

Distilling Knowledge from Language Models for Video-based Action Anticipation

Anticipating future actions in a video is useful for many autonomous and assistive technologies. Prior action anticipation work mostly treats this as a vision modality problem, where the models learn the task information primarily from the video features in the target action anticipation datasets. In this work, we propose a method to make use of the text-modality that is available during the training, to bring in complementary information that is not present in the target action anticipation datasets. In particular, we leverage pre-trained language models to build a text-modality teacher that is able to predict future actions based on text labels of the past actions extracted from the input video. To further adapt the teacher to the target domain (cooking), we also pretrain the teacher on textual instructions from a recipes dataset (Recipe1M). Then, we distill the knowledge gained by the text-modality teacher into a vision-modality student to further improve it's performance. We empirically evaluate this simple cross-modal distillation strategy on two video datasets EGTEA-GAZE+ and EPIC-KITCHEN 55. Distilling this text-modality knowledge into a strong vision model (Anticipative Vision Transformer) yields consistent gains across both datasets, 3.5% relative improvement on top1 class mean recall for EGTEA-GAZE+, 7.2% on top5 many-shot class mean recall for EPIC-KITCHEN 55 and achieves new state-of-the-results.

PASTA: A Dataset for Modeling Participant States in Narratives

The events in a narrative can be understood as a coherent whole via the underlying states of its participants. Often, these participant states are not explicitly mentioned in the narrative, left to be filled in via common-sense or inference. A model that understands narratives should be able to infer these implicit participant states and reason about the impact of changes to these states on the narrative. To facilitate this goal, we introduce a new crowdsourced Participants States dataset, PASTA. This dataset contains valid, inferable participant states; a counterfactual perturbation to the state; and the changes to the story that would be necessary if the counterfactual was true. We introduce three state-based reasoning tasks that test for the ability to infer when a state is entailed by a story, revise a story for a counterfactual state, and to explain the most likely state change given a revised story. Our benchmarking experiments show that while today's LLMs are able to reason about states to some degree, there is a large room for improvement, suggesting potential avenues for future research.

Teaching Broad Reasoning Skills via Decomposition-Guided Contexts

Question-answering datasets require a broad set of reasoning skills. We show how to use question decompositions to teach language models these broad reasoning skills in a robust fashion. Specifically, we use widely available QDMR representations to programmatically create synthetic contexts for real questions in six multihop reasoning datasets. These contexts are carefully designed to avoid common reasoning shortcuts prevalent in real contexts that prevent models from learning the right skills. This results in a pretraining dataset, named TeaBReaC, containing 525K multihop questions (with associated formal programs) covering about 900 reasoning patterns. We show that pretraining standard language models (LMs) on TeaBReaC before fine-tuning them on target datasets improves their performance by up to 13 EM points across 3 multihop QA datasets, with a 30 point gain on more complex questions. The resulting models also demonstrate higher robustness, with a 6-11 point improvement on two contrast sets. Furthermore, TeaBReaC pretraining substantially improves model performance and robustness even when starting with numeracy-aware LMs pretrained using recent methods (e.g., PReasM). Our work thus shows how one can effectively use decomposition-guided contexts to robustly teach multihop reasoning.