Recent studies show that Large Language Models (LLMs) achieve strong reasoning capabilities through supervised fine-tuning or reinforcement learning. However, a key approach, the Process Reward Model (PRM), suffers from reward hacking, making it unreliable in identifying the best intermediate step. In this paper, we propose a novel reward model approach, Hierarchical Reward Model (HRM), which evaluates both individual and consecutive reasoning steps from fine-grained and coarse-grained level. HRM excels in evaluating multi-step reasoning coherence and self-reflection, especially in scenarios where the previous reasoning step is incorrect but the subsequent step successfully identifies and corrects the error. Furthermore, to address the inefficiency of autonomous annotating PRM training data via Monte Carlo Tree Search (MCTS), we introduce a lightweight and effective data augmentation strategy called Hierarchical Node Compression (HNC) based on node merging (combining two consecutive reasoning steps into one step) in the tree structure. By applying HNC to MCTS-generated reasoning trajectories, we enhance the diversity and robustness of HRM training data while introducing controlled noise with minimal computational overhead. Empirical results on the PRM800K dataset demonstrate that HRM, in conjunction with HNC, achieves superior stability and reliability in evaluation compared to PRM. Furthermore, cross-domain evaluations on MATH500 and GSM8K dataset confirm HRM's superior generalization and robustness across diverse reasoning tasks.

Existing benchmarks often highlight the remarkable performance achieved by state-of-the-art Multimodal Foundation Models (MFMs) in leveraging temporal context for video understanding. However, how well do the models truly perform visual temporal reasoning? Our study of existing benchmarks shows that this capability of MFMs is likely overestimated as many questions can be solved by using a single, few, or out-of-order frames. To systematically examine current visual temporal reasoning tasks, we propose three principles with corresponding metrics: (1) Multi-Frame Gain, (2) Frame Order Sensitivity, and (3) Frame Information Disparity. Following these principles, we introduce TOMATO, Temporal Reasoning Multimodal Evaluation, a novel benchmark crafted to rigorously assess MFMs' temporal reasoning capabilities in video understanding. TOMATO comprises 1,484 carefully curated, human-annotated questions spanning six tasks (i.e., action count, direction, rotation, shape & trend, velocity & frequency, and visual cues), applied to 1,417 videos, including 805 self-recorded and -generated videos, that encompass human-centric, real-world, and simulated scenarios. Our comprehensive evaluation reveals a human-model performance gap of 57.3% with the best-performing model. Moreover, our in-depth analysis uncovers more fundamental limitations beyond this gap in current MFMs. While they can accurately recognize events in isolated frames, they fail to interpret these frames as a continuous sequence. We believe TOMATO will serve as a crucial testbed for evaluating the next-generation MFMs and as a call to the community to develop AI systems capable of comprehending human world dynamics through the video modality.

Prompting a pretrained language model with natural language patterns has been proved effective for natural language understanding (NLU). However, our preliminary study reveals that manual discrete prompts often lead to unstable performance -- e.g., changing a single word in the prompt might result in substantial performance drop. We propose a novel method P-Tuning that employs trainable continuous prompt embeddings in concatenation with discrete prompts. Empirically, P-Tuning not only stabilizes training by minimizing the gap between various discrete prompts, but also improves performance by a sizeable margin on a wide range of NLU tasks including LAMA and SuperGLUE. P-Tuning is generally effective for both frozen and tuned language models, under both the fully-supervised and few-shot settings.

Generative modeling has been the dominant approach for large-scale pretraining and zero-shot generalization. In this work, we challenge this convention by showing that discriminative approaches perform substantially better than generative ones on a large number of NLP tasks. Technically, we train a single discriminator to predict whether a text sample comes from the true data distribution, similar to GANs. Since many NLP tasks can be formulated as selecting from a few options, we use this discriminator to predict the concatenation of input and which option has the highest probability of coming from the true data distribution. This simple formulation achieves state-of-the-art zero-shot results on the T0 benchmark, outperforming T0 by 16.0\%, 7.8\%, and 11.5\% respectively on different scales. In the finetuning setting, our approach also achieves new state-of-the-art results on a wide range of NLP tasks, with only 1/4 parameters of previous methods. Meanwhile, our approach requires minimal prompting efforts, which largely improves robustness and is essential for real-world applications. Furthermore, we also jointly train a generalized UD in combination with generative tasks, which maintains its advantage on discriminative tasks and simultaneously works on generative tasks.

Natural language prompts have been shown to facilitate cross-task generalization for large language models. However, with no or limited labeled examples, the cross-task performance is highly sensitive to the choice of prompts, while selecting a high-performing prompt is challenging given the scarcity of labels. To address the issue, we propose a Zero-Label Prompt Selection (ZPS) method that selects prompts without any labeled data or gradient update. Specifically, given the candidate human-written prompts for a task, ZPS labels a set of unlabeled data with a prompt ensemble and uses the pseudo-labels for prompt selection. Experiments show that ZPS improves over prior methods by a sizeable margin in zero-label performance. We also extend ZPS to a few-shot setting and show its advantages over strong baselines such as prompt tuning and model tuning. Recently, extensive studies have shown that large language models (LLMs) have promising performance for few-shot learning (Brown et al., 2020; Zhao et al., 2021; Schick & Schütze, 2021; Gao et al., 2021), and they even show strong generalization abilities to new tasks without any annotated data (Brown et al., 2020; Wei et al., 2021; Sanh et al., 2021). Different from conventional fine-tuning methods that require expensive parameter updates for each downstream task, prompts are employed to provide in-context information or task instructions, which is helpful for guiding models to perform each task. Manually-written prompts are often used to specify the task and unify the format of inputs. However, the performance of different prompts during evaluation can vary from near state-of-the-art to random guess; e.g., using a non-optimal prompt can cause a performance drop of up to 60 points on the CB task (Zhao et al., 2021).

Few-shot named entity recognition (NER) targets generalizing to unseen labels and/or domains with few labeled examples. Existing metric learning methods compute token-level similarities between query and support sets, but are not able to fully incorporate label semantics into modeling. To address this issue, we propose a simple method to largely improve metric learning for NER: 1) multiple prompt schemas are designed to enhance label semantics; 2) we propose a novel architecture to effectively combine multiple prompt-based representations. Empirically, our method achieves new state-of-the-art (SOTA) results under 16 of the 18 considered settings, substantially outperforming the previous SOTA by an average of 8.84% and a maximum of 34.51% in relative gains of micro F1. Named entity recognition (NER) is a key natural language understanding task that extracts and classifies named entities mentioned in unstructured texts into predefined categories. Few-shot NER targets generalizing to unseen categories by learning from few labeled examples. Recent advances for few-shot NER use metric learning methods which compute the token-level similarities between the query and the given support cases. Snell et al. (2017) proposed to use prototypical networks that learn prototypical representations for target classes. Later, this method was introduced to few-shot NER tasks (Fritzler et al., 2019; Hou et al., 2020).