Safety evaluation of multimodal foundation models often treats vision and language inputs separately, missing risks from joint interpretation where benign content becomes harmful in combination. Existing approaches also fail to distinguish clearly unsafe content from borderline cases, leading to problematic over-blocking or under-refusal of genuinely harmful content. We present Vision Language Safety Understanding (VLSU), a comprehensive framework to systematically evaluate multimodal safety through fine-grained severity classification and combinatorial analysis across 17 distinct safety patterns. Using a multi-stage pipeline with real-world images and human annotation, we construct a large-scale benchmark of 8,187 samples spanning 15 harm categories. Our evaluation of eleven state-of-the-art models reveals systematic joint understanding failures: while models achieve 90%-plus accuracy on clear unimodal safety signals, performance degrades substantially to 20-55% when joint image-text reasoning is required to determine the safety label. Most critically, 34% of errors in joint image-text safety classification occur despite correct classification of the individual modalities, further demonstrating absent compositional reasoning capabilities. Additionally, we find that models struggle to balance refusing unsafe content while still responding to borderline cases that deserve engagement. For example, we find that instruction framing can reduce the over-blocking rate on borderline content from 62.4% to 10.4% in Gemini-1.5, but only at the cost of under-refusing on unsafe content with refusal rate dropping from 90.8% to 53.9%. Overall, our framework exposes weaknesses in joint image-text understanding and alignment gaps in current models, and provides a critical test bed to enable the next milestones in research on robust vision-language safety.

Prompt optimization has become crucial for enhancing the performance of large language models (LLMs) across a broad range of tasks. Although many research papers show its effectiveness, practical adoption is hindered as existing implementations are often tied to unmaintained and isolated research codebases. To address this, we introduce promptolution, a unified and modular open-source framework that provides all components required for prompt optimization within a single extensible system for both practitioners and researchers. It integrates multiple contemporary discrete prompt optimizers while remaining agnostic to the underlying LLM implementation.

In this paper, we report the results of the TeamNRC's participation in the BHASHA-Task 1 Grammatical Error Correction shared task https://github.com/BHASHA-Workshop/IndicGEC2025/ for 5 Indian languages. Our approach, focusing on zero/few-shot prompting of language models of varying sizes (4B to large proprietary models) achieved a Rank 4 in Telugu and Rank 2 in Hindi with GLEU scores of 83.78 and 84.31 respectively. In this paper, we extend the experiments to the other three languages of the shared task - Tamil, Malayalam and Bangla, and take a closer look at the data quality and evaluation metric used. Our results primarily highlight the potential of small language models, and summarize the concerns related to creating good quality datasets and appropriate metrics for this task that are suitable for Indian language scripts.

We propose and test the LLM Brain Rot Hypothesis: continual exposure to junk web text induces lasting cognitive decline in large language models (LLMs). To causally isolate data quality, we run controlled experiments on real Twitter/X corpora, constructing junk and reversely controlled datasets via two orthogonal operationalizations: M1 (engagement degree) and M2 (semantic quality), with matched token scale and training operations across conditions. Contrary to the control group, continual pre-training of 4 LLMs on the junk dataset causes non-trivial declines (Hedges' $g>0.3$) on reasoning, long-context understanding, safety, and inflating "dark traits" (e.g., psychopathy, narcissism). The gradual mixtures of junk and control datasets also yield dose-response cognition decay: for example, under M1, ARC-Challenge with Chain Of Thoughts drops $74.9 \rightarrow 57.2$ and RULER-CWE $84.4 \rightarrow 52.3$ as junk ratio rises from $0\%$ to $100\%$. Error forensics reveal several key insights. First, we identify thought-skipping as the primary lesion: models increasingly truncate or skip reasoning chains, explaining most of the error growth. Second, partial but incomplete healing is observed: scaling instruction tuning and clean data pre-training improve the declined cognition yet cannot restore baseline capability, suggesting persistent representational drift rather than format mismatch. Finally, we discover that the popularity, a non-semantic metric, of a tweet is a better indicator of the Brain Rot effect than the length in M1. Together, the results provide significant, multi-perspective evidence that data quality is a causal driver of LLM capability decay, reframing curation for continual pretraining as a \textit{training-time safety} problem and motivating routine "cognitive health checks" for deployed LLMs.

Reliably composing Large Language Models (LLMs) for complex, multi-step workflows remains a significant challenge. The dominant paradigm -- optimizing discrete prompts in a pipeline -- is notoriously brittle and struggles to enforce the formal compliance required for structured tasks. We introduce Type-Compliant Adaptation Cascades (TACs), a framework that recasts workflow adaptation as learning typed probabilistic programs. TACs treat the entire workflow, which is composed of parameter-efficiently adapted LLMs and deterministic logic, as an unnormalized joint distribution. This enables principled, gradient-based training even with latent intermediate structures. We provide theoretical justification for our tractable optimization objective, proving that the optimization bias vanishes as the model learns type compliance. Empirically, TACs significantly outperform state-of-the-art prompt-optimization baselines. Gains are particularly pronounced on structured tasks, improving FinQA from $12.0\%$ to $24.7\%$ for a Qwen 3 8B model, MGSM-SymPy from $57.1\%$ to $75.9\%$ for a Gemma 2 27B model, MGSM from $1.6\%$ to $27.3\%$, and MuSR from $36.5\%$ to $62.6\%$ for a Gemma 7B model. TACs offer a robust and theoretically grounded paradigm for developing reliable, task-compliant LLM systems.

Large language models (LLMs) offer significant potential to accelerate systematic literature reviews (SLRs), yet current approaches often rely on brittle, manually crafted prompts that compromise reliability and reproducibility. This fragility undermines scientific confidence in LLM-assisted evidence synthesis. In response, this work adapts recent advances in declarative prompt optimisation, developed for general-purpose LLM applications, and demonstrates their applicability to the domain of SLR automation. This research proposes a structured, domain-specific framework that embeds task declarations, test suites, and automated prompt tuning into a reproducible SLR workflow. These emerging methods are translated into a concrete blueprint with working code examples, enabling researchers to construct verifiable LLM pipelines that align with established principles of transparency and rigour in evidence synthesis. This is a novel application of such approaches to SLR pipelines.

Group Relative Policy Optimization (GRPO) has proven to be an effective tool for post-training language models (LMs). However, AI systems are increasingly expressed as modular programs that mix together multiple LM calls with distinct prompt templates and other tools, and it is not clear how best to leverage GRPO to improve these systems. We begin to address this challenge by defining mmGRPO, a simple multi-module generalization of GRPO that groups LM calls by module across rollouts and handles variable-length and interrupted trajectories. We find that mmGRPO, composed with automatic prompt optimization, improves accuracy by 11% on average across classification, many-hop search, and privacy-preserving delegation tasks against the post-trained LM, and by 5% against prompt optimization on its own. We open-source mmGRPO in DSPy as the dspy.GRPO optimizer.

Although prompt engineering is central to unlocking the full potential of Large Language Models (LLMs), crafting effective prompts remains a time-consuming trial-and-error process that relies on human intuition. This study investigates Declarative Self-improving Python (DSPy), an optimization framework that programmatically creates and refines prompts, applied to five use cases: guardrail enforcement, hallucination detection in code, code generation, routing agents, and prompt evaluation. Each use case explores how prompt optimization via DSPy influences performance. While some cases demonstrated modest improvements - such as minor gains in the guardrails use case and selective enhancements in hallucination detection - others showed notable benefits. The prompt evaluation criterion task demonstrated a substantial performance increase, rising accuracy from 46.2% to 64.0%. In the router agent case, the possibility of improving a poorly performing prompt and of a smaller model matching a stronger one through optimized prompting was explored. Although prompt refinement increased accuracy from 85.0% to 90.0%, using the optimized prompt with a cheaper model did not improve performance. Overall, this study's findings suggest that DSPy's systematic prompt optimization can enhance LLM performance, particularly when instruction tuning and example selection are optimized together. However, the impact varies by task, highlighting the importance of evaluating specific use cases in prompt optimization research.

Generating Wikipedia articles autonomously is a challenging task requiring the integration of accurate, comprehensive, and well-structured information from diverse sources. This paper introduces the Memory Organization-based Generation (MOG) framework, a novel approach to address these challenges by leveraging a hierarchical memory architecture. MOG extracts fine-grained memory units from web documents, recursively organizes them into a Wikipedia-style hierarchical structure, and uses this structure to guide the generation process. This ensures alignment between memory and the article outline, improving both informativeness and verifiability while minimizing hallucinations. Additionally, a citation module is implemented to enhance traceability by linking every generated sentence to specific memory units. Evaluations on our newly created WikiStart dataset demonstrate that MOG outperforms baseline methods in producing informative and reliable articles, making it particularly robust in real-world scenarios.

Large Language Models (LLMs) have reshaped natural language processing, powering applications from multi-hop retrieval and question answering to autonomous agent workflows. Yet, prompt engineering -- the task of crafting textual inputs to effectively direct LLMs -- remains difficult and labor-intensive, particularly for complex pipelines that combine multiple LLM calls with functional operations like retrieval and data formatting. We introduce LLM-AutoDiff: a novel framework for Automatic Prompt Engineering (APE) that extends textual gradient-based methods (such as Text-Grad) to multi-component, potentially cyclic LLM architectures. Implemented within the AdalFlow library, LLM-AutoDiff treats each textual input as a trainable parameter and uses a frozen backward engine LLM to generate feedback-akin to textual gradients -- that guide iterative prompt updates. Unlike prior single-node approaches, LLM-AutoDiff inherently accommodates functional nodes, preserves time-sequential behavior in repeated calls (e.g., multi-hop loops), and combats the "lost-in-the-middle" problem by isolating distinct sub-prompts (instructions, formats, or few-shot examples). It further boosts training efficiency by focusing on error-prone samples through selective gradient computation. Across diverse tasks, including single-step classification, multi-hop retrieval-based QA, and agent-driven pipelines, LLM-AutoDiff consistently outperforms existing textual gradient baselines in both accuracy and training cost. By unifying prompt optimization through a graph-centric lens, LLM-AutoDiff offers a powerful new paradigm for scaling and automating LLM workflows - mirroring the transformative role that automatic differentiation libraries have long played in neural network research.