Self-Explore to Avoid the Pit: Improving the Reasoning Capabilities of Language Models with Fine-grained Rewards

Training on large amounts of rationales (i.e., CoT Fine-tuning) is effective at improving the reasoning capabilities of large language models (LLMs). However, acquiring human-authored rationales or augmenting rationales from proprietary models is costly and not scalable. In this paper, we study the problem of whether LLMs could self-improve their reasoning capabilities. To this end, we propose Self-Explore, where the LLM is tasked to explore the first wrong step (i.e., the first pit) within the rationale and use such signals as fine-grained rewards for further improvement. On the GSM8K and MATH test set, Self-Explore achieves 11.57% and 2.89% improvement on average across three LLMs compared to supervised fine-tuning (SFT). Our code is available at https://github.com/hbin0701/Self-Explore.

Semiparametric Token-Sequence Co-Supervision

In this work, we introduce a semiparametric token-sequence co-supervision training method. It trains a language model by simultaneously leveraging supervision from the traditional next token prediction loss which is calculated over the parametric token embedding space and the next sequence prediction loss which is calculated over the nonparametric sequence embedding space. The nonparametric sequence embedding space is constructed by a separate language model tasked to condense an input text into a single representative embedding. Our experiments demonstrate that a model trained via both supervisions consistently surpasses models trained via each supervision independently. Analysis suggests that this co-supervision encourages a broader generalization capability across the model. Especially, the robustness of parametric token space which is established during the pretraining step tends to effectively enhance the stability of nonparametric sequence embedding space, a new space established by another language model.

Joint Mechanical and Electrical Adjustment of IRS-aided LEO Satellite MIMO Communications

In this letter, we propose a joint mechanical and electrical adjustment of intelligent reflecting surface (IRS) for the performance improvements of low-earth orbit (LEO) satellite multiple-input multiple-output (MIMO) communications. In particular, we construct a three-dimensional (3D) MIMO channel model for the mechanically-tilted IRS, and consider two types of scenarios with and without the direct path of LEO-ground user link due to the orbital flight. With the aim of maximizing the end-to-end performance, we jointly optimize tilting angle and phase shift of IRS along with the transceiver beamforming, whose performance superiority is verified via simulations.

Adaptive Shortcut Debiasing for Online Continual Learning

We propose a novel framework DropTop that suppresses the shortcut bias in online continual learning (OCL) while being adaptive to the varying degree of the shortcut bias incurred by continuously changing environment. By the observed high-attention property of the shortcut bias, highly-activated features are considered candidates for debiasing. More importantly, resolving the limitation of the online environment where prior knowledge and auxiliary data are not ready, two novel techniques -- feature map fusion and adaptive intensity shifting -- enable us to automatically determine the appropriate level and proportion of the candidate shortcut features to be dropped. Extensive experiments on five benchmark datasets demonstrate that, when combined with various OCL algorithms, DropTop increases the average accuracy by up to 10.4% and decreases the forgetting by up to 63.2%.

One Size Fits All for Semantic Shifts: Adaptive Prompt Tuning for Continual Learning

In real-world continual learning scenarios, tasks often exhibit intricate and unpredictable semantic shifts, posing challenges for fixed prompt management strategies. We identify the inadequacy of universal and specific prompting in handling these dynamic shifts. Universal prompting is ineffective for tasks with abrupt semantic changes, while specific prompting struggles with overfitting under mild semantic shifts. To overcome these limitations, we propose an adaptive prompting approach that tailors minimal yet sufficient prompts based on the task semantics. Our methodology, SemPrompt, incorporates a two-level semantic grouping process: macroscopic semantic assignment and microscopic semantic refinement. This process ensures optimal prompt utilization for varying task semantics, improving the efficiency and effectiveness of learning in real-world CL settings. Our experimental results demonstrate that SemPrompt consistently outperforms existing methods in adapting to diverse semantic shifts in tasks.

How Well Do Large Language Models Truly Ground?

Reliance on the inherent knowledge of Large Language Models (LLMs) can cause issues such as hallucinations, lack of control, and difficulties in integrating variable knowledge. To mitigate this, LLMs can be probed to generate responses by grounding on external context, often given as input (knowledge-augmented models). Yet, previous research is often confined to a narrow view of the term "grounding", often only focusing on whether the response contains the correct answer or not, which does not ensure the reliability of the entire response. To address this limitation, we introduce a strict definition of grounding: a model is considered truly grounded when its responses (1) fully utilize necessary knowledge from the provided context, and (2) don't exceed the knowledge within the contexts. We introduce a new dataset and a grounding metric to assess this new definition and perform experiments across 13 LLMs of different sizes and training methods to provide insights into the factors that influence grounding performance. Our findings contribute to a better understanding of how to improve grounding capabilities and suggest an area of improvement toward more reliable and controllable LLM applications.

Robust Data Pruning under Label Noise via Maximizing Re-labeling Accuracy

Data pruning, which aims to downsize a large training set into a small informative subset, is crucial for reducing the enormous computational costs of modern deep learning. Though large-scale data collections invariably contain annotation noise and numerous robust learning methods have been developed, data pruning for the noise-robust learning scenario has received little attention. With state-of-the-art Re-labeling methods that self-correct erroneous labels while training, it is challenging to identify which subset induces the most accurate re-labeling of erroneous labels in the entire training set. In this paper, we formalize the problem of data pruning with re-labeling. We first show that the likelihood of a training example being correctly re-labeled is proportional to the prediction confidence of its neighborhood in the subset. Therefore, we propose a novel data pruning algorithm, Prune4Rel, that finds a subset maximizing the total neighborhood confidence of all training examples, thereby maximizing the re-labeling accuracy and generalization performance. Extensive experiments on four real and one synthetic noisy datasets show that \algname{} outperforms the baselines with Re-labeling models by up to 9.1% as well as those with a standard model by up to 21.6%.

Energy-Efficient Secure Offloading System Designed via UAV-Mounted Intelligent Reflecting Surface for Resilience Enhancement

With increasing interest in mmWave and THz communication systems, an unmanned aerial vehicle (UAV)-mounted intelligent reflecting surface (IRS) has been suggested as a key enabling technology to establish robust line-of-sight (LoS) connections with ground nodes owing to their free mobility and high altitude, especially for emergency and disaster response. This paper investigates a secure offloading system, where the UAV-mounted IRS assists the offloading procedures between ground users and an access point (AP) acting as an edge cloud. In this system, the users except the intended recipients in the offloading process are considered as potential eavesdroppers. The system aims to achieve the minimum total energy consumption of battery-limited ground user devices under constraints for secure offloading accomplishment and operability of UAV-mounted IRS, which is done by optimizing the transmit power of ground user devices, the trajectory and phase shift matrix of UAV-mounted IRS, and the offloading ratio between local execution and edge computing based on the successive convex approximation (SCA) algorithms. Numerical results show that the proposed algorithm can provide the considerable energy savings compared with local execution and partial optimizations.

FLASK: Fine-grained Language Model Evaluation based on Alignment Skill Sets

Evaluation of Large Language Models (LLMs) is challenging because aligning to human values requires the composition of multiple skills and the required set of skills varies depending on the instruction. Recent studies have evaluated the performance of LLMs in two ways, (1) automatic evaluation on several independent benchmarks and (2) human or machined-based evaluation giving an overall score to the response. However, both settings are coarse-grained evaluations, not considering the nature of user instructions that require instance-wise skill composition, which limits the interpretation of the true capabilities of LLMs. In this paper, we introduce FLASK (Fine-grained Language Model Evaluation based on Alignment SKill Sets), a fine-grained evaluation protocol that can be used for both model-based and human-based evaluation which decomposes coarse-level scoring to an instance-wise skill set-level. Specifically, we define 12 fine-grained skills needed for LLMs to follow open-ended user instructions and construct an evaluation set by allocating a set of skills for each instance. Additionally, by annotating the target domains and difficulty level for each instance, FLASK provides a holistic view with a comprehensive analysis of a model's performance depending on skill, domain, and difficulty. Through using FLASK, we compare multiple open-sourced and proprietary LLMs and observe highly-correlated findings between model-based and human-based evaluations. FLASK enables developers to more accurately measure the model performance and how it can be improved by analyzing factors that make LLMs proficient in particular skills. For practitioners, FLASK can be used to recommend suitable models for particular situations through comprehensive comparison among various LLMs. We release the evaluation data and code implementation at https://github.com/kaistAI/FLASK.

The CoT Collection: Improving Zero-shot and Few-shot Learning of Language Models via Chain-of-Thought Fine-Tuning

Large Language Models (LLMs) have shown enhanced capabilities of solving novel tasks by reasoning step-by-step known as Chain-of-Thought (CoT) reasoning; how can we instill the same capability of reasoning step-by-step on unseen tasks into LMs that possess less than <100B parameters? To address this question, we first introduce the CoT Collection, a new instruction-tuning dataset that augments 1.88 million CoT rationales across 1,060 tasks. We show that continually fine-tuning Flan-T5 (3B & 11B) with the CoT Collection enables the 3B & 11B LMs to perform CoT better on unseen tasks, leading to an improvement in the average zero-shot accuracy on 27 datasets of the BIG-Bench-Hard benchmark by +4.34% and +2.44%, respectively. Furthermore, we show that instruction tuning with CoT allows LMs to possess stronger few-shot learning capabilities, resulting in an improvement of +2.97% and +2.37% on 4 domain-specific tasks over Flan-T5 (3B & 11B), respectively. We make our CoT Collection data and our trained models publicly available at https://github.com/kaist-lklab/CoT-Collection.