When Text and Numbers Disagree: Evidence Arbitration in Large Language Models

Large language models (LLMs) are increasingly used in settings where textual summaries, numerical observations, and external tool outputs may provide conflicting evidence. We study how LLMs arbitrate between such sources when they support opposing decisions. To do so, we introduce a controlled synthetic benchmark in which latent risk trajectories generate both numerical time series and natural language summaries, allowing us to construct conflicts where exactly one evidence source is aligned with the ground-truth label. This design lets us independently manipulate modality, temporal recency, source reliability, and evidence provenance. Across open-weight instruction-tuned models, we find that arbitration behaviour is systematic rather than random: models exhibit distinct text-versus-number preferences, follow temporal recency more consistently than explicit reliability cues, and can over-rely on external forecasts even when they conflict with direct contextual evidence. These results suggest that current LLMs often rely on heuristic arbitration strategies when integrating heterogeneous evidence, highlighting a failure mode for tool-augmented decision systems.

When Do Explanations Help In-Context Learning? A Comparative Study of Natural Language Explanation Types and Faithfulness

Natural language explanations (NLEs) are increasingly used as inputs, for example, as few-shot rationales that influence model behavior in in-context learning (ICL). However, it remains unclear how different types of NLEs compare in their effects on downstream model performance in explanation-augmented prompting. Therefore, we provide a comparative evaluation across six benchmarks and four instruction-tuned models, studying how NLE source (human-written when available, self-generated explanations, generated by an external LLM) and NLE selection (random vs faithfulness-based filtering) affect downstream utility of NLEs when used in ICL settings. Our extensive evaluation shows that, on classification-style benchmarks, adding NLEs to few-shot prompts often improves accuracy over few-shot prompting without explanations; among NLE sources, externally generated LLM-NLEs often provide strong downstream utility and remain competitive with human rationales where both are available, whereas self-NLEs are more sensitive to the selection strategy. On math reasoning, the effects are more model- and source-dependent. We further show that faithfulness-based selection of self-NLEs yields small average gains overall, but can improve or reduce performance depending on the metric, task, and model. Different faithfulness metrics can disagree substantially, affecting which self-NLE examples are selected and their downstream predictive utility. Robustness tests with randomly swapped and out-of-distribution rationales indicate partial robustness, suggesting that semantic alignment contributes to performance gains. Overall, our results provide insights for selecting and reporting explanations that influence model behavior in practical prompting pipelines.

Where Steering Signals Come From: Activation Source Selection in Activation Steering

Activation steering controls language models by adding vectors or features to hidden states at inference time, but the upstream source of these steering signals is often treated as a secondary detail. We study this source choice as activation source selection: the combination of source context and activation readout policy used to collect the hidden states from which a steering signal is built. Holding the downstream intervention fixed, we show across three instruction-tuned models and four steering task families that changing only the source activations substantially changes steering success. We further find that effective steering is not explained simply by whether the desired behavior appears in the source text. Instead, strong signals come from execution-boundary states, where the model is about to produce or continue the target behavior. This pre-/post-realization distinction explains why answer-based sources sometimes work: their useful component aligns with execution-boundary directions rather than target appearance alone. Building on this view, we introduce tail subtraction, which removes shared prompt and continuation semantics from boundary states and yields cleaner, more stable steering signals. Overall, our results suggest that steering depends on representations of what the model is about to do, not merely on what has already appeared.

Discovering Millions of Interpretable Features with Sparse Autoencoders

Sparse autoencoders (SAEs) have emerged as a powerful tool for decomposing superposed language model representations into sparse and interpretable features. However, training SAEs is computationally expensive, and available open-source SAE models remain limited. In this work, we introduce \textbf{Qwen3-Instruct SAE}, a comprehensive suite of SAEs trained on the Qwen3 instruction-tuned model family, covering Qwen3-1.7B, Qwen3-4B, and Qwen3-8B. For Qwen3-1.7B and Qwen3-4B, we train layer-wise SAEs at three key activation sites: residual streams, MLP outputs, and attention outputs. For Qwen3-8B, we train SAEs on a subset of residual stream layers. We systematically evaluate these SAEs using both activation-level reconstruction metrics and model-level recovery metrics, revealing distinct sparsity--fidelity trade-offs across layers and components. Finally, we demonstrate the utility of Qwen3-Instruct SAE through a refusal-steering case study, showing that selected SAE features can causally steer instruction-tuned Qwen3 models toward refusal behavior. Our release provides a practical resource for studying sparse representations, feature-level mechanisms, and behavioral interventions in instruction-tuned language models

When Retrieval Doesn't Help: A Large-Scale Study of Biomedical RAG

Medical question answering is a high-stakes setting where factual errors can have serious consequences. Retrieval-augmented generation (RAG) is widely viewed as a promising solution, and prior work has reported substantial gains for large medical QA models. We revisit this assumption across a broad range of open-weight instruction-tuned models spanning 7B to 72B parameters. Across five models, ten biomedical QA datasets, four retrieval methods, and four retrieval corpora, we find that retrieval yields only small and inconsistent improvements over a no-retrieval baseline, typically within 1-2 points. In contrast, the choice of backbone model has a much larger effect than the choice of retriever or corpus, and expert and layman retrieval sources perform similarly in most settings. These results suggest that the main bottleneck is not retrieval quality alone, but the model's limited ability to use retrieved evidence effectively.