Named after the hundred-eyed watchman of Greek myth, Argus watches the education landscape: spotting new opportunities, pressure-testing the ventures we're building, and tracing every read back to the real-world signals behind it.
Signals [67] and [4] confirm that epistemic stance is routinely lost during LLM memory compression and that chain-of-thought reasoning is unreliable as an explanation mechanism; signal [17] shows long reasoning chains actively distort LLM judgment of factuality; signal [64] demonstrates LLMs systematically reinforce user biases; and signal [44] shows internal activation-based fact-checking is now technically viable—together making a real-time epistemic integrity wrapper both necessary and buildable.
When students use AI writing or research assistants, the tools strip epistemic qualifiers (uncertainty hedges, confidence levels, source attribution) from generated content during summarization and memory compression, producing outputs that mislead students about what is known versus inferred.
Higher education institutions, K-12 districts deploying AI writing assistants, and ed-tech platforms building on top of LLM APIs for student research support.
A middleware layer that wraps existing LLM-based writing and research tools to detect, preserve, and visually surface epistemic stance markers throughout summarization, retrieval, and generation. It flags where AI-generated claims lack grounding, where compression has dropped hedging language, and where the model's stated confidence is unreliable—delivering inline source-traceable annotations that teach students to read AI output critically while protecting academic integrity.
The real-world evidence the pipeline drew on to generate this idea.
arXiv:2608.06953v1 Announce Type: new Abstract: Agent memory systems compress what they store, and compression is built to drop qualifiers, so a claim's epistemic standing tends not to survive being written to memory. We ask what governs whether it does. Matched notes carry the identical claim and identical stance and differ only in where that stance sits; one model compresses both under the same budget among the same filler notes, and a blind reader that never sees the condition scores the result. Across 60 claims in seven registers, writing the stance as a labelled field rather than a bracketed aside raises retention by about 15 points on two models (37 claims to 2 on one, 30 to 8 on the other; permutation p=0.00005), and a pre-registered replication on Haiku, its prediction and decision rule committed before the run, gives +15.6 points, 38 claims to 1. Ablating the format on both models gives the same net effect from different parts: labels help on both (+9.7 and +12.8) and length h
arXiv:2602.20710v2 Announce Type: replace-cross Abstract: Inspecting Chain-of-Thought reasoning is among the most common means of understanding why an LLM produced its output. But well-known problems with CoT faithfulness severely limit what insights can be gained from this practice. In this paper, we introduce a training method called Counterfactual Simulation Training (CST), which aims to improve CoT faithfulness by rewarding CoTs that enable a simulator to accurately predict a model's outputs over counterfactual inputs. We apply CST in two settings: (1) CoT monitoring with cue-based counterfactuals, to detect when models rely on spurious features, reward hack, or are sycophantic, and (2) counterfactual simulation over generic model-based counterfactuals, to encourage models to produce more faithful, generalizable reasoning in the CoT. Experiments with models up to 235B parameters show that CST can substantially improve monitor accuracy on cue-based counterfactuals (by 35 accuracy po
arXiv:2604.06756v2 Announce Type: replace Abstract: Large language models (LLMs) has been widely adopted as a scalable surrogate for human evaluation, yet such judges remain imperfect and susceptible to surface-level biases. One possible reason is that these judges lack sufficient information in assessing answer correctness. With the rise of reasoning-capable models, exposing a generator's reasoning content to the judge provides richer information and is a natural candidate for improving judgment accuracy. However, its actual impact on judge behavior remains understudied. In this paper, we systematically investigate how access to reasoning chains affects LLM-based judgment across factual question answering (QA) and mathematical reasoning benchmarks. We find that weak judges are easily swayed by reasoning presence, frequently accepting incorrect answers accompanied by fluent reasoning, while strong judges can partially leverage reasoning as informative evidence. Nevertheless, even stron
arXiv:2608.06977v1 Announce Type: new Abstract: It is well established that large language models (LLMs) are sensitive to prompt framing, reflecting patterns in their training data or prior prompts. In this study, we investigate the extent to which LLMs reinforce users biases expressed in the prompts and examine the boundary between implicit framing effects and explicit prompt manipulation. Specifically, we evaluate how susceptible LLMs are to direct and suggestive prompts that encourage models to support or challenge particular positions. We evaluate six LLMs using 160 distinct prompts spanning ten topics across opinion-based and factual domains. The prompts systematically vary in prompting strategy, support versus challenge instructions, prompt polarity, users' expressed beliefs, and topic domain, spanning both opinion-based and factual questions. Our results show that LLMs systematically adapt their responses to align with prompt framing, even in factual contexts. This suggests that
arXiv:2608.06417v1 Announce Type: cross Abstract: The proliferation of misinformation online has driven demand for scalable detection systems. While most existing approaches rely on surface-level linguistic features or external knowledge retrieval, we examine truthfulness as a geometric property of a language model's representation space. We introduce a misinformation detection framework grounded in activation engineering, which leverages the latent geometry of transformer models. Our approach elicits a misinformation direction in the residual stream by contrasting activations from paired truthful and false statements, following the difference-in-means principle of Contrastive Activation Addition (CAA). At inference time, the last-token activation of an unseen claim is projected onto this direction, and the projected representation is fed to an Multilayer Perceptron (MLP) for classification. The procedure requires no fine-tuning of the backbone model, no external evidence retrieval, an
arXiv:2505.16170v4 Announce Type: replace Abstract: We study the internal mechanisms that govern when LLMs choose to retract wrong answers, i.e., spontaneously and immediately acknowledge errors in their previously generated false assertions. Using model-specific testbeds, we find that while LLMs are capable of retraction, they do so only rarely, even when they can recognize their mistakes when asked in a separate interaction. We identify a reliable predictor of retraction: the model's momentary belief, as measured by a linear probe on its internal representation. The probe is trained to predict the correctness of answers on external datasets unrelated to retraction, then applied to settings where models should retract. A model retracts only when it "believes" its answers to be incorrect during generation; these beliefs frequently diverge from models' parametric knowledge as measured by factoid questions. Steering experiments further demonstrate that model belief causally drives retrac
arXiv:2608.07370v1 Announce Type: new Abstract: Scientific literature is increasingly used as a knowledge source for language models, retrieval-augmented generation systems, and research assistants, but answering research questions from papers requires more than fluent generation. A reliable system must identify the relevant papers, locate the concrete evidence that supports the answer, and produce a response that is faithful to that evidence. We present LitTraceQA, a benchmark for literature-grounded question answering over scientific papers. Given a research question and a metadata pool of papers, a system must return three connected outputs: canonical paper identifiers, supporting evidence locations, and answers in one or more requested formats, including free-form text, multiple-choice answers, and structured tables. LitTraceQA targets evidence types common in scientific reading: tables, figures, text spans, equations or algorithms, and citation contexts. The public development spl