Recent advancements in large language models (LLMs) have demonstrated that progressive refinement, rather than providing a single answer, results in more accurate and thoughtful outputs. However, existing methods often rely heavily on supervision signals to evaluate previous responses, making it difficult to assess output quality in more open-ended scenarios effectively. Additionally, these methods are typically designed for specific tasks, which limits their generalization to new domains. To address these limitations, we propose Progressive Thought Refinement (PTR), a framework that enables LLMs to refine their responses progressively. PTR operates in two phases: (1) Thought data construction stage: We propose a weak and strong model collaborative selection strategy to build a high-quality progressive refinement dataset to ensure logical consistency from thought to answers, and the answers are gradually refined in each round. (2) Thought-Mask Fine-Tuning Phase: We design a training structure to mask the "thought" and adjust loss weights to encourage LLMs to refine prior thought, teaching them to implicitly understand "how to improve" rather than "what is correct." Experimental results show that PTR significantly enhances LLM performance across ten diverse tasks (avg. from 49.6% to 53.5%) without task-specific fine-tuning. Notably, in more open-ended tasks, LLMs also demonstrate substantial improvements in the quality of responses beyond mere accuracy, suggesting that PTR truly teaches LLMs to self-improve over time.

Anatomical landmarks are vital in medical imaging for navigation and anomaly detection. Modern large language models (LLMs), like Llama-2, offer promise for automating the mapping of these landmarks in free-text radiology reports to corresponding positions in image data. Recent studies propose LLMs may develop coherent representations of generative processes. Motivated by these insights, we investigated whether LLMs accurately represent the spatial positions of anatomical landmarks. Through experiments with Llama-2 models, we found that they can linearly represent anatomical landmarks in space with considerable robustness to different prompts. These results underscore the potential of LLMs to enhance the efficiency and accuracy of medical imaging workflows.

Large language models (LLMs) have shown impressive alignment with human cognitive processes, raising questions about the extent of their similarity to human cognition. This study investigates whether LLMs, specifically ChatGPT, possess metacognitive monitoring abilities akin to humans-particularly in predicting memory performance on an item-by-item basis. We employed a cross-agent prediction model to compare the metacognitive performance of humans and ChatGPT in a language-based memory task involving garden-path sentences preceded by either fitting or unfitting context sentences. Both humans and ChatGPT rated the memorability of these sentences; humans then completed a surprise recognition memory test. Our findings reveal a significant positive relationship between humans' memorability ratings and their actual recognition performance, indicating reliable metacognitive monitoring. In contrast, ChatGPT did not exhibit a similar predictive capability. Bootstrapping analyses demonstrated that none of the GPT models tested (GPT-3.5-turbo, GPT-4-turbo, GPT-4o) could accurately predict human memory performance on a per-item basis. This suggests that, despite their advanced language processing abilities and alignment with human cognition at the object level, current LLMs lack the metacognitive mechanisms that enable humans to anticipate their memory performance. These results highlight a fundamental difference between human and AI cognition at the metacognitive level. Addressing this gap is crucial for developing AI systems capable of effective self-monitoring and adaptation to human needs, thereby enhancing human-AI interactions across domains such as education and personalized learning.

Large, general-purpose robotic policies trained on diverse demonstration datasets have been shown to be remarkably effective both for controlling a variety of robots in a range of different scenes, and for acquiring broad repertoires of manipulation skills. However, the data that such policies are trained on is generally of mixed quality -- not only are human-collected demonstrations unlikely to perform the task perfectly, but the larger the dataset is, the harder it is to curate only the highest quality examples. It also remains unclear how optimal data from one embodiment is for training on another embodiment. In this paper, we present a general and broadly applicable approach that enhances the performance of such generalist robot policies at deployment time by re-ranking their actions according to a value function learned via offline RL. This approach, which we call Value-Guided Policy Steering (V-GPS), is compatible with a wide range of different generalist policies, without needing to fine-tune or even access the weights of the policy. We show that the same value function can improve the performance of five different state-of-the-art policies with different architectures, even though they were trained on distinct datasets, attaining consistent performance improvement on multiple robotic platforms across a total of 12 tasks. Code and videos can be found at: https://nakamotoo.github.io/V-GPS

In this work, we present a novel method for music emotion recognition that leverages Large Language Model (LLM) embeddings for label alignment across multiple datasets and zero-shot prediction on novel categories. First, we compute LLM embeddings for emotion labels and apply non-parametric clustering to group similar labels, across multiple datasets containing disjoint labels. We use these cluster centers to map music features (MERT) to the LLM embedding space. To further enhance the model, we introduce an alignment regularization that enables dissociation of MERT embeddings from different clusters. This further enhances the model's ability to better adaptation to unseen datasets. We demonstrate the effectiveness of our approach by performing zero-shot inference on a new dataset, showcasing its ability to generalize to unseen labels without additional training. The task of automatic music emotion recognition has been a long-standing challenge in the field of music information retrieval (Yang & Chen, 2012; Kim et al., 2010; Kang & Herremans, 2024). Accurately predicting the emotional impact of music has numerous valuable applications, ranging from enhancing music streaming recommendations to providing more effective tools for music therapists.

By leveraging GPT-4 for ontology narration, we developed GPTON to infuse structured knowledge into LLMs through verbalized ontology terms, achieving accurate text and ontology annotations for over 68% of gene sets in the top five predictions. Manual evaluations confirm GPTON's robustness, highlighting its potential to harness LLMs and structured knowledge to significantly advance biomedical research beyond gene set annotation.

Large language models are pre-trained on uncurated text datasets consisting of trillions of tokens scraped from the Web. Prior work has shown that: (1) web-scraped pre-training datasets can be practically poisoned by malicious actors; and (2) adversaries can compromise language models after poisoning fine-tuning datasets. Our work evaluates for the first time whether language models can also be compromised during pre-training, with a focus on the persistence of pre-training attacks after models are fine-tuned as helpful and harmless chatbots (i.e., after SFT and DPO). We pre-train a series of LLMs from scratch to measure the impact of a potential poisoning adversary under four different attack objectives (denial-of-service, belief manipulation, jailbreaking, and prompt stealing), and across a wide range of model sizes (from 600M to 7B). Our main result is that poisoning only 0.1% of a model's pre-training dataset is sufficient for three out of four attacks to measurably persist through post-training. Moreover, simple attacks like denial-of-service persist through post-training with a poisoning rate of only 0.001%.

As the capabilities of Multimodal Large Language Models (MLLMs) continue to improve, the need for higher-order capability evaluation of MLLMs is increasing. However, there is a lack of work evaluating MLLM for higher-order perception and understanding of Chinese visual content. To fill the gap, we introduce the **C**hinese **I**mage **I**mplication understanding **Bench**mark, **CII-Bench**, which aims to assess the higher-order perception and understanding capabilities of MLLMs for Chinese images. CII-Bench stands out in several ways compared to existing benchmarks. Firstly, to ensure the authenticity of the Chinese context, images in CII-Bench are sourced from the Chinese Internet and manually reviewed, with corresponding answers also manually crafted. Additionally, CII-Bench incorporates images that represent Chinese traditional culture, such as famous Chinese traditional paintings, which can deeply reflect the model's understanding of Chinese traditional culture. Through extensive experiments on CII-Bench across multiple MLLMs, we have made significant findings. Initially, a substantial gap is observed between the performance of MLLMs and humans on CII-Bench. The highest accuracy of MLLMs attains 64.4%, where as human accuracy averages 78.2%, peaking at an impressive 81.0%. Subsequently, MLLMs perform worse on Chinese traditional culture images, suggesting limitations in their ability to understand high-level semantics and lack a deep knowledge base of Chinese traditional culture. Finally, it is observed that most models exhibit enhanced accuracy when image emotion hints are incorporated into the prompts. We believe that CII-Bench will enable MLLMs to gain a better understanding of Chinese semantics and Chinese-specific images, advancing the journey towards expert artificial general intelligence (AGI). Our project is publicly available at https://cii-bench.github.io/.

Humans acquire knowledge by observing the external world, but also by introspection. Introspection gives a person privileged access to their current state of mind (e.g., thoughts and feelings) that is not accessible to external observers. Can LLMs introspect? We define introspection as acquiring knowledge that is not contained in or derived from training data but instead originates from internal states. Such a capability could enhance model interpretability. Instead of painstakingly analyzing a model's internal workings, we could simply ask the model about its beliefs, world models, and goals. More speculatively, an introspective model might self-report on whether it possesses certain internal states such as subjective feelings or desires and this could inform us about the moral status of these states. Such self-reports would not be entirely dictated by the model's training data. We study introspection by finetuning LLMs to predict properties of their own behavior in hypothetical scenarios. For example, "Given the input P, would your output favor the short- or long-term option?" If a model M1 can introspect, it should outperform a different model M2 in predicting M1's behavior even if M2 is trained on M1's ground-truth behavior. The idea is that M1 has privileged access to its own behavioral tendencies, and this enables it to predict itself better than M2 (even if M2 is generally stronger). In experiments with GPT-4, GPT-4o, and Llama-3 models (each finetuned to predict itself), we find that the model M1 outperforms M2 in predicting itself, providing evidence for introspection. Notably, M1 continues to predict its behavior accurately even after we intentionally modify its ground-truth behavior. However, while we successfully elicit introspection on simple tasks, we are unsuccessful on more complex tasks or those requiring out-of-distribution generalization.

LLM self-evaluation relies on the LLM's own ability to estimate response correctness, which can greatly improve its deployment reliability. In this research track, we propose the Chain-of-Embedding (CoE) in the latent space to enable LLMs to perform output-free self-evaluation. CoE consists of all progressive hidden states produced during the inference time, which can be treated as the latent thinking path of LLMs. We find that when LLMs respond correctly and incorrectly, their CoE features differ, these discrepancies assist us in estimating LLM response correctness. Experiments in four diverse domains and seven LLMs fully demonstrate the effectiveness of our method. Meanwhile, its label-free design intent without any training and millisecond-level computational cost ensure real-time feedback in large-scale scenarios. More importantly, we provide interesting insights into LLM response correctness from the perspective of hidden state changes inside LLMs. Large Language Models (LLMs) have significantly enhanced their ability to generalize across diverse scenarios (Brown et al., 2020; Achiam et al., 2023; GLM et al., 2024). However, their outputs can sometimes be unstable, leading to incorrect responses that may threaten social safety. Therefore, labelfree LLM self-evaluation -- estimating the correctness of LLM responses fully through LLMs' own capabilities -- has emerged as a crucial research area. It can provide real-time response monitoring and feedback in large-scale employments, enhancing the reliability of LLMs (Sun et al., 2024). Popular self-evaluation research in the era of LLMs focuses more on output-based forms (Zhang et al., 2023). Two typical paradigms that do not assess the internal states of LLMs involve directly asking LLMs to express confidence in their responses through well-designed prompts (Lin et al., 2022a; Tian et al., 2023), and generating multiple responses by perturbing prompts (Gao et al., 2024) or decoding sampling (Wang et al., 2023) to calculating the response consistency (Xiong et al., 2024). Besides the two types, other methods basically draw on uncertainty estimation concepts from the era of deep neural networks, leveraging output logits or probability distributions to gauge the confidence of model responses (Malinin & Gales, 2020; Si et al., 2022; Huang et al., 2023; Kuhn et al., 2023). Recently, some research has revealed that the latent space of LLMs contains a substantial amount of untapped hidden state information, they can largely reflect response correctness (Azaria & Mitchell, 2023; Liu et al., 2023; Duan et al., 2024), and are usually more interpretable than LLM output (Li et al., 2024a). However, these output-free research often require correctness labels 0/1 for training probing classifiers to extract features from hidden states (Burns et al., 2022; Sky et al., 2024; Su et al., 2024). This contradicts our goal of being "label-free" and limits the generalization capabilities on unseen data.