Evaluating image captions requires cohesive assessment of both visual semantics and language pragmatics, which is often not entirely captured by most metrics. We introduce Redemption Score(RS), a novel hybrid framework that ranks image captions by triangulating three complementary signals: (1) Mutual Information Divergence (MID) for global image-text distributional alignment, (2) DINO-based perceptual similarity of cycle-generated images for visual grounding, and (3) LLM Text Embeddings for contextual text similarity against human references. A calibrated fusion of these signals allows RS to offer a more holistic assessment. On the Flickr8k benchmark, RS achieves a Kendall-$τ$ of 58.42, outperforming most prior methods and demonstrating superior correlation with human judgments without requiring task-specific training. Our framework provides a more robust and nuanced evaluation by thoroughly examining both the visual accuracy and text quality together, with consistent performance across Conceptual Captions and MS COCO.

Modern BPE tokenizers often split calendar dates into meaningless fragments, e.g., 20250312 $\rightarrow$ 202, 503, 12, inflating token counts and obscuring the inherent structure needed for robust temporal reasoning. In this work, we (1) introduce a simple yet interpretable metric, termed date fragmentation ratio, that measures how faithfully a tokenizer preserves multi-digit date components; (2) release DateAugBench, a suite of 6500 examples spanning three temporal reasoning tasks: context-based date resolution, format-invariance puzzles, and date arithmetic across historical, contemporary, and future time periods; and (3) through layer-wise probing and causal attention-hop analyses, uncover an emergent date-abstraction mechanism whereby large language models stitch together the fragments of month, day, and year components for temporal reasoning. Our experiments show that excessive fragmentation correlates with accuracy drops of up to 10 points on uncommon dates like historical and futuristic dates. Further, we find that the larger the model, the faster the emergent date abstraction that heals date fragments is accomplished. Lastly, we observe a reasoning path that LLMs follow to assemble date fragments, typically differing from human interpretation (year $\rightarrow$ month $\rightarrow$ day). Our datasets and code are made publicly available \href{https://github.com/gagan3012/date-fragments}{here}.

Reinforcement learning (RL) has emerged as an effective approach for enhancing the reasoning capabilities of large language models (LLMs), especially in scenarios where supervised fine-tuning (SFT) falls short due to limited chain-of-thought (CoT) data. Among RL-based post-training methods, group relative advantage estimation, as exemplified by Group Relative Policy Optimization (GRPO), has attracted considerable attention for eliminating the dependency on the value model, thereby simplifying training compared to traditional approaches like Proximal Policy Optimization (PPO). However, we observe that existing group relative advantage estimation method still suffers from training inefficiencies, particularly when the estimated advantage approaches zero. To address this limitation, we propose Advantage-Augmented Policy Optimization (AAPO), a novel RL algorithm that optimizes the cross-entropy (CE) loss using advantages enhanced through a momentum-based estimation scheme. This approach effectively mitigates the inefficiencies associated with group relative advantage estimation. Experimental results on multiple mathematical reasoning benchmarks and model series demonstrate the superior performance of AAPO. Reinforcement learning (RL) has emerged as a powerful approach for enhancing the reasoning and decision-making capabilities of large language models (LLMs). To address this limitation, recent research has explored RL-based training paradigms, which have shown considerable empirical success in specialized domains such as mathematical reasoning.

Large Language Models (LLMs) are increasingly applied in various science domains, yet their broader adoption remains constrained by a critical challenge: the lack of trustworthy, verifiable outputs. Current LLMs often generate answers without reliable source attribution, or worse, with incorrect attributions, posing a barrier to their use in scientific and high-stakes settings, where traceability and accountability are paramount. To be reliable, attribution systems require high accuracy for short-length attribution on retrieved data, i.e., attribution to a sentence within a document rather than the entire document. We propose SAFE, a Sentence-level A ttribution FramEwork for Retrieve-Augmented Generation (RAG) systems that attributes generated sentences during generation. This allows users to verify sentences as they read them and correct the model when the attribution indicates the generated text is not grounded in the documents, increasing the safety of LLM systems. This framework consists of two steps: predicting the required number of references for a sentence, and attributing the sentence. Our approach achieved 95% accuracy in the first step, which translated to 2.1\~6.0% improvements in the accuracy (normalized for maximum possible accuracy) of all attribution algorithms in our clean dataset, when compared to their top-1 accuracy. We also applied SAFE in real-world scenarios with documents containing hundreds to thousands of sentences. In these settings, SAFE reliably attributed sentences to their source documents, demonstrating that the method generalizes beyond controlled benchmarks. The SAFE framework and the training dataset are publicly available on GitHub.

Augmenting large language models (LLMs) with external retrieval has become a standard method to address their inherent knowledge cutoff limitations. However, traditional retrieval-augmented generation methods employ static, pre-inference retrieval strategies, making them inadequate for complex tasks involving ambiguous, multi-step, or evolving information needs. Recent advances in test-time scaling techniques have demonstrated significant potential in enabling LLMs to dynamically interact with external tools, motivating the shift toward adaptive inference-time retrieval. Inspired by Information Foraging Theory (IFT), we propose InForage, a reinforcement learning framework that formalizes retrieval-augmented reasoning as a dynamic information-seeking process. Unlike existing approaches, InForage explicitly rewards intermediate retrieval quality, encouraging LLMs to iteratively gather and integrate information through adaptive search behaviors. To facilitate training, we construct a human-guided dataset capturing iterative search and reasoning trajectories for complex, real-world web tasks. Extensive evaluations across general question answering, multi-hop reasoning tasks, and a newly developed real-time web QA dataset demonstrate InForage's superior performance over baseline methods. These results highlight InForage's effectiveness in building robust, adaptive, and efficient reasoning agents.

The advantage function is a central concept in RL that helps reduce variance in policy gradient estimates. Recently, for language modeling, Group Relative Policy Optimization (GRPO) was proposed to compute the advantage for each output by subtracting the mean reward, as the baseline, for all outputs in the group. However, it can lead to high variance when the reward advantage is inaccurately predicted. In this work, we propose Kalman Filter Enhanced Group Relative Policy Optimization (KRPO) model, by using lightweight Kalman filtering to dynamically estimate the latent reward baseline and uncertainty. This filtering technique replaces the naive group mean, enabling more adaptive advantage normalization. Our method does not require additional learned parameters over GRPO. This approach offers a simple yet effective way to incorporate multiple outputs of GRPO into advantage estimation, improving policy optimization in settings where highly dynamic reward signals are difficult to model for language models. Through the accuracies and rewards obtained from math question answering and reasoning, we show that using a more adaptive advantage estimation model, KRPO can improve the stability and performance of GRPO. The code is available at https://github.com/billhhh/KRPO_LLMs_RL.

This study aims to guide language model selection by investigating: 1) the necessity of finetuning versus zero-shot usage, 2) the benefits of domain-adjacent versus generic pretrained models, 3) the value of further domain-specific pretraining, and 4) the continued relevance of Small Language Models (SLMs) compared to Large Language Models (LLMs) for specific tasks. Using electronic pathology reports from the British Columbia Cancer Registry (BCCR), three classification scenarios with varying difficulty and data size are evaluated. Models include various SLMs and an LLM. SLMs are evaluated both zero-shot and finetuned; the LLM is evaluated zero-shot only. Finetuning significantly improved SLM performance across all scenarios compared to their zero-shot results. The zero-shot LLM outperformed zero-shot SLMs but was consistently outperformed by finetuned SLMs. Domain-adjacent SLMs generally performed better than the generic SLM after finetuning, especially on harder tasks. Further domain-specific pretraining yielded modest gains on easier tasks but significant improvements on the complex, data-scarce task. The results highlight the critical role of finetuning for SLMs in specialized domains, enabling them to surpass zero-shot LLM performance on targeted classification tasks. Pretraining on domain-adjacent or domain-specific data provides further advantages, particularly for complex problems or limited finetuning data. While LLMs offer strong zero-shot capabilities, their performance on these specific tasks did not match that of appropriately finetuned SLMs. In the era of LLMs, SLMs remain relevant and effective, offering a potentially superior performance-resource trade-off compared to LLMs.

Despite the existing evolution of Multimodal Large Language Models (MLLMs), a non-neglectable limitation remains in their struggle with visual text grounding, especially in text-rich images of documents. Document images, such as scanned forms and infographics, highlight critical challenges due to their complex layouts and textual content. However, current benchmarks do not fully address these challenges, as they mostly focus on visual grounding on natural images, rather than text-rich document images. Thus, to bridge this gap, we introduce TRIG, a novel task with a newly designed instruction dataset for benchmarking and improving the Text-Rich Image Grounding capabilities of MLLMs in document question-answering. Specifically, we propose an OCR-LLM-human interaction pipeline to create 800 manually annotated question-answer pairs as a benchmark and a large-scale training set of 90$ synthetic data based on four diverse datasets. A comprehensive evaluation of various MLLMs on our proposed benchmark exposes substantial limitations in their grounding capability on text-rich images. In addition, we propose two simple and effective TRIG methods based on general instruction tuning and plug-and-play efficient embedding, respectively. By finetuning MLLMs on our synthetic dataset, they promisingly improve spatial reasoning and grounding capabilities.

Reinforcement learning with verifiable rewards (RL VR) improves language model reasoning by using rule-based rewards in verifiable domains such as mathematics and code. However, RL VR leads to limited generalization for open-ended tasks--such as writing outline essays or making meal plans--where humans reason routinely. This paper shows that the RL VR paradigm is effective beyond verifiable domains, and introduces RL with Model-rewarded Thinking (RLMT) for general-purpose chat capabilities. Using diverse real-world prompts, RLMT requires LMs to generate long CoT reasoning before response, and optimizes them with online RL against a preference-based reward model used in RLHF. Across 40 training runs on Llama-3.1-8B and Qwen-2.5-7B This includes substantial gains of 3-7 points on three chat benchmarks (AlpacaEval2, WildBench, and Arena-HardV2), along with 1-3 point improvements on other tasks like creative writing and general knowledge. Our best 8B model surpasses GPT -4o in chat and creative writing and rivals Claude-3.7-Sonnet RLMT can also be applied directly to base models without an SFT stage, akin to R1-Zero training (DeepSeek-AI, 2025). Remarkably, with only 7K prompts, Llama-3.1-8B We close with qualitative and quantitative analyses of how trained models plan their responses. Our results rethink the post-training pipeline and call upon future work to understand and employ thinking more broadly. HINKING through the consequences of one's actions--and revising them when needed--is a defining feature of human intelligence (often called "system 2 thinking", Kahneman (2011)). It has also become a central aspiration for large language models (LLMs).

Learning visuomotor policies from expert demonstrations is an important frontier in modern robotics research, however, most popular methods require copious efforts for collecting teleoperation data and struggle to generalize out-ofdistribution. Scaling data collection has been explored through leveraging human videos, as well as demonstration augmentation techniques. The latter approach typically requires expensive simulation rollouts and trains policies with synthetic image data, therefore introducing a sim-to-real gap. In parallel, alternative state representations such as keypoints have shown great promise for category-level generalization. In this work, we bring these avenues together in a unified framework: PAD (Parse-AugmentDistill), for learning generalizable bimanual policies from a single human video. Our method relies on three steps: (a) parsing a human video demo into a robot-executable keypoint-action trajectory, (b) employing bimanual task-and-motion-planning to augment the demonstration at scale without simulators, and (c) distilling the augmented trajectories into a keypoint-conditioned policy. Empirically, we showcase that PAD outperforms state-ofthe-art bimanual demonstration augmentation works relying on image policies with simulation rollouts, both in terms of success rate and sample/cost efficiency. We deploy our framework in six diverse real-world bimanual tasks such as pouring drinks, cleaning trash and opening containers, producing one-shot policies that generalize in unseen spatial arrangements, object instances and background distractors. Supplementary material can be found in the project webpage https://gtziafas.github.io/PAD_project/.