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.
The evidence library: the raw signals the pipeline is watching across the education ecosystem. Every idea is built from these.
While the U.S. celebrates over the Fourth of July weekend, four candidates will be vying for the top post of the nation’s largest teachers union. The National Education Association’s leadership election will decide the replacement for President Becky Pringle and other officers during the union’s annual representative assembly from July 3 to 7 in Denver. […]
The new law took effect immediately and came after legislators overrode a veto from Democratic Gov. Josh Stein.
New York City education officials are hitting pause on releasing comprehensive artificial intelligence guidelines after their draft policy from March sparked fierce backlash. Officials initially said their final guidance would be released in June, but backed away from that timeline during a Wednesday City Council hearing focused on AI in schools. Instead, the policy guidance […]
More than 9 in 10 U.S. school districts say they have planned to purchase Chromebooks for their students or staff this year. So, when rumors began to buzz about the Googlebook, which brings together elements of ChromeOS and Android, many district leaders understandably had a lot of questions. Will my school’s devices eventually become obsolete? What can I do to future proof my device program? We have good news for K–12 districts that are concerned about this change: The real story is that schools have ample time and multiple options to prepare for the upcoming upgrades. By beginning to plan…
Educators and advocates woory that reduced public school funding and a lack of resources has taken its toll already. But there is hope that opportunities exist, if the state can get behind them. The post Ohio public school districts face teacher shortages amid reduced state funding, budget woes appeared first on District Administration .
The proposal, which mandates literary works such as Charles Dickens' "Great Expectations" alongside passages from the New Testament, has been closely followed by education observers who say it appears to be the first of its kind in the nation. The post Texas education board approves Bible stories as required reading in public schools appeared first on District Administration .
Article URL: https://metis-learn.io/ Comments URL: https://news.ycombinator.com/item?id=48719104 Points: 2 # Comments: 0
Stephania Zamorano has been an educator in New York City for 15 years. Since 2021, she has worked at Imagine Early Learning Centers, a childcare organization that serves nearly 600 children across 12 sites, soon to be 13, in the New York metropolitan area. Zamorano described her time at Imagine as quite different from her […]
PETERSBURG, Va. — For years, the schools here have been stuck in a very bad place. As Petersburg’s once-booming manufacturing base hollowed out, crime increased and residents’ health worsened. Along the way, places for kids to be kids disappeared, and many of them stopped coming to school. Over the last year, about a quarter of […]
A new mobile simulation lab at Sutter Health means more access to high quality immersive training for nurses in diverse locations, says this CNO. As high quality patient care continues to be paramount for health systems, high quality training for nurses remains a goal for CNOs. However, when time and money are precious and workflows are busy, it can be difficult to find the right balance of standardized training across a health system. According to Kat Ascencio-Holmes , system CNO at Sutter Health , one of the biggest hurdles that CNOs face when standardizing training is ensuring equitable access to learning opportunities across geographically diverse locations. "Another is creating a consistency in that training while supporting the unique needs of the individual communities that we serve," Ascencio-Holmes said. That’s why leadership at Sutter Health implemented a new, first-of-its-kind, mobile simulation lab. The lab will address both of those challenges, Ascencio-Holmes explained. "
Students are more than letter grades and test scores. They are unique individuals with their own goals, skills, and needs--deserving an educational journey that supports all three.
System transformation rarely fails because of a bad strategy. It fails because the relationships needed to sustain it were never built. In this piece, Rebecca Midles and Nate McClennen draw on real district leaders to show what coalition-building actually looks like when the messy middle arrives. From a rural Michigan superintendent who started with the willing to a Kansas City microschool that put students in the design seat, this is a practical and deeply human look at what makes change last. The post Coalition Before Consensus: How Trust and Shared Ownership Sustain Transformation appeared first on Getting Smart .
A new study suggests that critical thinking can be taught when the right strategies are in place.
Smore is the interactive newsletter builder that's made for education.
As generative AI becomes a routine part of academic work, a familiar question keeps surfacing in classrooms and scholarly writing alike: What, exactly, should writers disclose? The post Transparency appendices may be the next essential AI disclosure practice in higher education appeared first on eCampus News .
Article URL: https://xg.glass/posts/network-exam-test/ Comments URL: https://news.ycombinator.com/item?id=48716055 Points: 2 # Comments: 2
The Apprenticeship Wish List Johanna Alonso Mon, 06/29/2026 - 03:00 AM Apprenticeships are growing. But experts say that without more funding, updated laws and better data, the U.S. is still far from the system they know is possible. Byline(s) Johanna Alonso
We Know the Ph.D. Job Market Is Broken Elizabeth Redden Mon, 06/29/2026 - 03:00 AM Do we know what could help? Byline(s) Dinuka Gunaratne Roger Pizarro Milian
Michigan Regent Apologizes Following Sexual Comments Josh Moody Mon, 06/29/2026 - 03:00 AM Byline(s) Josh Moody
Regional Public Students Feel Belonging, but Also Financial Stress Emma Whitford Mon, 06/29/2026 - 03:00 AM Byline(s) Emma Whitford
Delaware County Community College Moves to Cut Counseling Sara Weissman Mon, 06/29/2026 - 03:00 AM The community college issued layoff notices to all of its part-time and full-time counselors, citing budget concerns and possible plans to outsource services. Byline(s) Sara Weissman
Is Automation ‘Distorting’ the History of Scientific Research? kathryn.palmer… Mon, 06/29/2026 - 03:00 AM Byline(s) Kathryn Palmer
The Job Market for Recent College Grads in 5 Charts kathryn.palmer… Mon, 06/29/2026 - 03:00 AM An economist helped us cut through the noise to capture a clear, nuanced picture of what the labor market for new graduates really looks like right now. Byline(s) Kathryn Palmer
Free Tuition for Adult Community College Learners Joshua.Bay Mon, 06/29/2026 - 03:00 AM A Milken Institute report finds tuition support for adult learners at two-year institutions could improve retention, increase credential completion and boost earnings. Byline(s) Joshua Bay
The Researcher Who Spent Years Studying Grass for the World Cup Johanna Alonso Mon, 06/29/2026 - 03:00 AM John Sorochan runs the turfgrass science and management program at the University of Tennessee at Knoxville. In recent months, his work has gained national recognition for its contributions to the World Cup pitches. Byline(s) Johanna Alonso
‘What About Your Kids?’ Sara Brady Mon, 06/29/2026 - 03:00 AM Sniffing out hypocrisy when it comes to public college employees’ kids attending the college at which their parent works. Byline(s) Matt Reed
Article URL: https://www.bbc.co.uk/news/articles/cq51j10q601o Comments URL: https://news.ycombinator.com/item?id=48715260 Points: 3 # Comments: 0
For years, policymakers, educators and employers have debated whether career pathways — programs that connect high school students to postsecondary education and careers — actually work. We’ve framed the conversation as apprenticeship versus college, workforce training versus liberal arts and careers versus academics. While new findings from Rodel and RTI International — in one of […] The post OPINION: We need to ask better questions about how and if career pathways are working appeared first on The Hechinger Report .
Harrison Keller was starting only his second year as president of the University of North Texas last fall when he was abruptly confronted with a big problem. Enrollment was down. And the source of the crisis made it much worse: In the wake of Trump administration moves to deny and revoke visas, deport international students […] The post As international enrollment falls, U.S. students face program cuts and higher prices appeared first on The Hechinger Report .
The two federal agencies are offering eligible institutions up to $30 million for property improvements — but the money requires a one-to-one match.
We’re rounding up last week’s stories, from a record-high persistence rate to the latest federal lawsuit over in-state tuition for undocumented students.
Teacher burnout rates have remained high since COVID-19, but experts say artificial intelligence is still a promising solution if done right — and at scale.
We’re rounding up last week’s news, from calls for Linda McMahon’s impeachment to data on states not meeting special education requirements.
OAKLAND, Calif. — It’s a Friday morning at Oakland Unity Middle School, a public charter school nestled between residential buildings in East Oakland, and Austin Razavi is announcing the morning advisory prompt. “I’ll give you 10, 15 seconds to think about it,” Razavi said to the group of 15 mixed-grade middle school boys who had […] The post Key to helping boys in school: Make them feel safe to be themselves appeared first on The Hechinger Report .
arXiv:2606.16682v3 Announce Type: replace-cross Abstract: When AI agents use language models to evaluate their own outputs in a feedback loop, systematic biases emerge. We show that Evaluator Preference Collapse (EPC) is dramatically amplified in multimodal settings. Using GPT-4o to evaluate DeepSeek-chat across text and visual tasks, we find that a single strategy (step_by_step) absorbs 48.4% of all weight -- 3.2x the collapse observed in text-only self-evaluation -- while three visual-domain strategies receive only 9.1% combined weight. We then demonstrate a novel phenomenon we term cross-modal coupling: evaluator preferences acquired on one modality transfer to and corrupt strategy selection on another. Through a four-phase isolation training paradigm, we measure coupling coefficients and document strategy inversion -- the optimal strategy for a modality reverses after cross-modal exposure. A Phase 3 statistical validation across five evaluator configurations (N=80 total independent
arXiv:2606.13544v3 Announce Type: replace-cross Abstract: Turn-taking in multi-party spoken conversations remains a fundamental challenge for voice-based agents, particularly under dynamic floor competition and varying user expectations. We propose ModeratorLM, a role-playing voice agent that conditions turn-taking behavior on an explicitly assigned role in multi-party settings. The system is built on a speech large language model operating in chunk-wise streaming manner. We further introduce a reasoning-augmented variant that incorporates chain-of-thought reasoning over conversational context and the assigned role. We construct RolePlayConv, a large-scale synthetic dataset of spoken multi-party conversations with diverse assistant roles. Experiments on real-world meeting data and RolePlayConv show improved turn-taking precision by over 40% and recall by more than 70%, while substantially reducing false-positive interruptions compared to non-role-conditioned baselines.
arXiv:2606.11585v2 Announce Type: replace-cross Abstract: Transformer models are increasingly used as computational models of cognition and neural representation, so the mechanism implemented by self-attention is of interest beyond engineering performance. A complementary tradition in cognitive science models coordination, binding, and memory through dynamical interactions such as oscillator synchrony; we bring this mechanism into self-attention by introducing the Kuramoto Attention layer, whose value update is a synchronization step. Each token carries a bank of phase oscillators, so its hidden state lives on a high-dimensional torus. The attention weights form an adaptive coupling graph, and using the raw phase states as values makes the value update exactly the Kuramoto coupling direction for fixed attention weights. The softmax selects which oscillators couple, while the value path moves each token toward the attention-weighted circular mean of the tokens it selects. We train Kuram
arXiv:2606.10156v2 Announce Type: replace-cross Abstract: As recommender systems transition toward agentic, multi-turn conversational interfaces, evaluation paradigms have struggled to keep pace. Current benchmarks often rely on "LLM-as-a-judge" evaluations, which introduce subjectivity, high costs and inconsistency. We present $\tau$-Rec, a benchmark for agentic recommender systems that replaces subjective evaluation with verifiable rewards and a reveal-tagged elicitation (RTE) mechanism that controls how task constraints surface during dialogue. By testing agents against structured catalog predicates and employing a pass^k reliability metric, $\tau$-Rec provides a systematic test for consistent reasoning. Our evaluation of nine configurations across five model families -- GPT-5.4, Claude Sonnet 4.6, Gemini 2.5 Flash, DeepSeek V4 Flash, Qwen3-32B and GPT-5 mini -- reveals a steep reliability cliff, where even the best model achieves only ~57% at pass^1 and ~35% at pass^4, highlighting
arXiv:2606.09774v2 Announce Type: replace-cross Abstract: Configuring an advanced scientific simulator, translating a modeling goal into a valid, runnable input deck, is a persistent bottleneck that costs domain scientists hours to days. Input decks are executable interfaces: simulator-specific vocabulary, cross-file references, schema constraints, and validation rules must align before a simulation can run. We show that this bottleneck can be substantially reduced with a lightweight adapter around an off-the-shelf coding agent, rather than a bespoke simulator agent. Coding agents already navigate files, edit code, run commands, and repair outputs; what they lack is the simulator's executable contract, and rebuilding the agent loop risks discarding harness-calibrated tool-use and self-correction behavior. We introduce SIGA, a coding-agent adapter that supplies this contract through retrieval, procedural memory, agent-callable validation, and validation-gated termination while leaving t
arXiv:2605.26872v3 Announce Type: replace-cross Abstract: LLM training increasingly relies on teacher-generated supervision, from synthetic responses to reasoning traces and tool-use demonstrations. Current practice often chooses the highest-performing teacher to generate student training data, implicitly treating teacher test performance as a proxy for teaching quality. We show that this assumption can fail: even when multiple teachers provide correct answers to the same question, the answer from the strongest teacher is not necessarily the best supervision for a given student. To address this gap, we propose Student-Centric Answer Sampling (SCAS), a framework that selects from verified teacher-generated answers according to their estimated student-centric learning cost. Motivated by a token-wise gradient decomposition, we derive an efficient forward-only proxy for this cost and use it to guide answer selection during training. Experiments across 30 teacher models, 6 student base mode
arXiv:2605.22536v2 Announce Type: replace-cross Abstract: Multimodal Large Language Models (MLLMs) have made rapid progress in spatial intelligence, yet existing spatial reasoning benchmarks largely assume pristine visual inputs and overlook the degradations that commonly occur in real-world deployment, such as motion blur, low light, adverse weather, lens distortion, and compression artifacts. This raises a fundamental question: how robust is the spatial intelligence of current MLLMs when visual observations are imperfect? To answer this question, we introduce SpaceDG, the first large-scale dataset for degradation-aware spatial understanding. It is constructed with a physically grounded degradation synthesis engine that embeds degradation formation process into 3D Gaussian Splatting (3DGS) rendering, enabling realistic simulation of nine degradation types. The resulting dataset contains approximately 1M QA pairs from nearly 1,000 indoor scenes. We further introduce SpaceDG-Bench, an h
arXiv:2605.14568v3 Announce Type: replace-cross Abstract: Context. Behaviour-Driven Development (BDD) test suites accumulate duplicated step subsequences. Three published refactoring patterns are available (within-file Background, within-repo reusable-scenario invocation, cross-organisational shared higher-level step), but no prior work automates which recurring subsequences are worth extracting or which mechanism applies. Objective. Rank recurring step subsequences ("slices") by refactoring suitability (extraction-worthy), pre-map each to one of the three patterns, and quantify prevalence across the public BDD ecosystem. Method. Every contiguous L-step window (L in [2, 18]) in a 339-repository / 276-upstream-owner Gherkin corpus is keyed by paraphrase-robust cluster identifiers and counted under three scopes. SBERT / UMAP / HDBSCAN clustering recovers paraphrase-equivalent slices. Three authors label a stratified 200-slice pool against a written rubric. An XGBoost extraction-worthy cl
arXiv:2605.06675v2 Announce Type: replace-cross Abstract: Large language models cache all previously computed key-value (KV) pairs during generation, and this KV cache grows linearly with sequence length, making it a primary memory bottleneck for serving. Quantizing the KV cache to fewer bits reduces this cost, yet all current quantizers assign the same bit-width to every attention head, ignoring the large variation in head importance. A natural idea is to allocate more bits to important heads and fewer to the rest. We show, however, that such mixed-precision allocation has a hidden pitfall: each quantizer follows a different distortion curve D(b)=alpha*beta^{-b}, and the decay rate beta varies from 3.6 to 5.3 across quantizer designs. Applying one quantizer's distortion model to another inverts the allocation order and makes performance worse than uniform quantization. We call this failure mode distortion model mismatch and propose RateQuant to resolve it. RateQuant fits a per-quantiz
arXiv:2603.09731v3 Announce Type: replace-cross Abstract: Multimodal large language models (MLLMs) are increasingly considered as a foundation for embodied agents, yet it remains unclear whether they can reliably reason about the long-term physical consequences of actions from an egocentric viewpoint. We study this gap through a new task, Egocentric Scene Prediction with LOng-horizon REasoning: given an initial-scene image and a sequence of atomic action descriptions, a model is asked to predict the final scene after all actions are executed. To enable systematic evaluation, we introduce EXPLORE-Bench, a benchmark curated from real first-person videos spanning diverse scenarios. Each instance pairs long action sequences with structured final-scene annotations, including object categories, visual attributes, and inter-object relations, which supports fine-grained, quantitative assessment. Experiments on a range of proprietary and open-source MLLMs reveal a significant performance gap to
arXiv:2510.18874v3 Announce Type: replace-cross Abstract: Adapting language models (LMs) to new tasks via post-training carries the risk of degrading existing capabilities -- a phenomenon classically known as catastrophic forgetting. In this paper, toward identifying guidelines for mitigating this phenomenon, we systematically compare the forgetting patterns of two widely adopted post-training methods: supervised fine-tuning (SFT) and reinforcement learning (RL). Our experiments reveal a consistent trend across LM families (Llama, Qwen) and tasks (instruction following, general knowledge, and arithmetic reasoning): RL leads to less forgetting than SFT while achieving comparable or higher target task performance. To investigate the cause for this difference, we consider a simplified setting in which the LM is modeled as a mixture of two distributions, one corresponding to prior knowledge and the other to the target task. We identify that the mode-seeking nature of RL, which stems from i
arXiv:2506.16150v4 Announce Type: replace-cross Abstract: As large language models (LLMs) advance, concerns about their misconduct in complex social contexts intensify. Existing research overlooked the systematic understanding and assessment of their criminal capability in realistic interactions. We propose a unified framework PRISON, to quantify LLMs' criminal potential across five traits: False Statements, Frame-Up, Psychological Manipulation, Emotional Disguise, and Moral Disengagement. Using structured crime scenarios adapted from classic films grounded in reality, we evaluate both criminal potential and anti-crime ability of LLMs. Results show that state-of-the-art LLMs frequently exhibit emergent criminal tendencies, such as proposing misleading statements or evasion tactics, even without explicit instructions. Moreover, when placed in a detective role, models recognize deceptive behavior with only 44% accuracy on average, revealing a striking mismatch between conducting and dete
arXiv:2506.07460v2 Announce Type: replace-cross Abstract: Sign language generation (SLG), also known as text-to-sign generation, aims to bridge the communication gap between signers and non-signers. Unlike many other generative tasks, SLG must satisfy two fundamental linguistic constraints. First, sign language expresses meaning through a sequence of gestures aligned with word-like units called glosses, and therefore requires correct lexical ordering to preserve intended meaning. Second, each gesture should faithfully reflect the intended gloss (semantic accuracy). Despite recent progress, existing SLG methods frequently produce signs with incorrect lexical order and low semantic accuracy. A common limitation of prior approaches stems from globally fused conditioning strategies, which weaken temporal grounding, the temporal correspondence between glosses and their realized sign segments. This often leads to incorrect lexical order and semantically ambiguous signs. To address this limit
arXiv:2606.26530v2 Announce Type: replace Abstract: The Abstraction and Reasoning Corpus (ARC) contains tasks that require summarizing patterns from limited grid samples and predicting output grids. Recently, many large language model based approaches have attempted to transform it into a text-based reasoning task. However, methods based on open-source models have generally yielded unsatisfactory results, while those relying on closed-source models are too costly. Current efforts mainly focus on data augmentation, constructing ARC-like data for more comprehensive supervised fine-tuning. In this work, we argue that solving ARC-like problems requires not only positive sample supervision but also the ability to improve model reasoning by distinguishing negative samples. To this end, we draw on the idea of preference alignment and propose DiARC, a method that constructs preference pairs to enable the model to distinguish between them. Specifically, we propose three ways to construct negati
arXiv:2606.23387v2 Announce Type: replace Abstract: Self-stigma predicts treatment avoidance and disengagement among people who use drugs (PWUD), yet conversational systems aiming to provide support typically treat self-stigma expression as a uniform signal. We present a three-phase, proof-of-concept study of a persona-aware approach to LLM support. Latent Profile Analysis (LPA) on indicator-level features from 1,174 self-stigma expressors on Reddit yields a four-persona typology validated against held-out behavioral and linguistic features. Sequential Bayesian and recurrent neural classifiers recover these personas from limited posting histories, substantially outperforming batch and few-shot LLM baselines (macro-F1 = 0.74 at 30 posts). Evaluation by eight clinical experts across three contemporary LLMs revealed a misalignment: persona-matched responses successfully achieved targeted behavioral shifts, yet raters holistically preferred the generic empathy of the persona-neutral baseli