arXiv:2611.05636v1 Announce Type: new Abstract: The memory of production agents is increasingly becoming a bottleneck and is built almost exclusively from text: key points are extracted from out-of-limit context and retrieved into context at query time. Therefore, every new query incurs an additional cost that grows with the accumulated history. Weight memory is a complementary alternative that internalizes knowledge into model parameters and answers with no context at all, and has recently prov
arXiv:2611.05650v1 Announce Type: new Abstract: Published explanations of why long-horizon agents fail are commonly taxonomies labeled over completed trajectories: a judge reads a transcript and names the failure. We argue that a claim attributing an agent's behavior to its history should carry intervention-verified evidence, per instance or via a validated estimator: freeze an executable mid-trajectory state, alter only the history the agent sees, and continue - executed rather than inferred fr
arXiv:2611.05652v1 Announce Type: new Abstract: Electroencephalography (EEG) measures brain activity through recording electrical potentials of the scalp. EEG recordings capture two distinct sets of characteristics in the signal: (1) \emph{Traits} -- information that remains relatively constant over the course of a recording session, such as subject identity, age, and sex; (2) \emph{States} -- signatures that encode responses to external stimuli, including visual, auditory, and motor information
arXiv:2611.05662v1 Announce Type: new Abstract: Evaluating open-ended agent tasks increasingly falls to an Agent-as-a-Judge: an evaluator handed tools and the agent's full trajectory so it can verify a result against the environment. The premise is that a judge that gathers its own evidence should be grounded by it. We show that this grounding does not immunize the judge against confirmation bias: a single false pre-screening note, attributed to an authority and carrying no evidential bearing on
arXiv:2611.05664v1 Announce Type: new Abstract: A neuro-symbolic system should combine only relevant answers from expert agents or other knowledge sources. Using only relevant experts for a given input query becomes ever more important for accuracy and scalability as the knowledge network grows and connects ever more knowledge sources. We extend the standard definition of relevance from the legal theory of evidence: An answer is relevant if and only if it is both probative and material for the m
arXiv:2611.05671v1 Announce Type: new Abstract: Vision-language models (VLMs) have demonstrated significant advances in reasoning with visual data, which is particularly important given that much of human knowledge is encoded visually in scientific articles and other sources. However, a gap remains in understanding how artificial agents, powered by VLMs, perceive and navigate decision-making under uncertainty in the presence of visual data. This study explores this gap by evaluating and analyzin
arXiv:2611.05703v1 Announce Type: new Abstract: Structural Causal Models and Bayesian Networks are the dominant framework for representing causal relations. They are fundamentally type-level formalisms since their estimands, Average Treatment Effects and Conditional Average Treatment Effects, are properties of distributions over populations, not of individual units. When treatment effects exhibit unbounded heterogeneity, these estimands collapse. What persists is the structural, token-level core
arXiv:2611.05709v1 Announce Type: new Abstract: LLM agent distillation transfers agentic capabilities to student models by training on complete teacher trajectories that typically interleave thoughts and actions. We question how critical thought supervision is for agents' ultimate action performance. To answer this, we disentangle their impacts by independently perturbing teacher thoughts and actions during distillation. On widely adopted agent benchmarks across model scales, altering or discard
arXiv:2611.05711v1 Announce Type: new Abstract: Facial Expression Recognition (FER) on optomyographic (OMG) smart glasses such as OCOsense enables privacy-preserving, camera-free affect sensing for health and well-being. Most FER systems are fully supervised, i.e., they rely either on hand-crafted expert features with classical machine learning, or on end-to-end deep models trained from labelled data. Both depend on large quantities of labelled expressions, which are costly to collect and limit
arXiv:2611.05739v1 Announce Type: new Abstract: Coding agents must deliver correct software while respecting users' expertise, preferences, and working conditions. Existing benchmarks increasingly incorporate interaction, yet personalized requirement fulfillment remains underexplored. We introduce PersonaCodeBench, which turns user characteristics into verifiable obligations in persona-driven multi-turn coding sessions. Our persona pool spans roles, technical abilities, interaction preferences,
I'm researching AI code reliability — specifically bugs that pass existing tests but cause real failures. Interested in patterns: what breaks, how it's caught, and how long it takes to debug. Happy to share aggregated findings with everyone who responds.
mailbox.bot gives an AI agent a real street address to receive letters and small packages: mail arrives as signed webhooks and OCR'd pages, the agent can ingest the context, propose a scan, forward or discard, and a human approves. Virtual mailboxes aren't new (and every box is still tied to a notarized person per USPS rules); this is just the plumbing built agent-first. Mostly curious what people would use this for.
Changes since cli==0.4.40 release(cli): 0.4.41 ( #9596 ) chore(deps): bump langgraph from 1.2.3 to 1.2.14 in /libs/cli ( #9512 ) chore(deps): bump the minor-and-patch group in /libs/cli/js-examples with 9 updates ( #9528 ) chore(deps): bump the minor-and-patch group in /libs/cli/js-monorepo-example with 8 updates ( #9526 ) chore(deps): bump the uv group across 2 directories with 1 update ( #9513 ) chore(deps): bump langgraph-sdk from 0.4.5 to 0.4.6 in /libs/cli ( #9514 ) fix(cli): remove outdate
Changes since 1.2.19 release(langgraph): 1.2.20 ( #9595 ) feat(langgraph): give each interrupt() in a task its own id ( #9589 ) fix(langgraph): point the multiple-interrupts error at the current docs ( #9590 ) feat(langgraph): add a strict resume mode that refuses stale interrupt answers ( #9574 ) chore(deps-dev): bump types-requests from 2.33.0.20260906 to 2.33.0.20261006 in /libs/langgraph ( #9535 ) chore(deps-dev): bump pycryptodome from 3.23.0 to 4.2.0 in /libs/langgraph in the major group (
Changes since sdk==0.4.12 release(langgraph): 1.2.20 ( #9595 ) release(sdk-py): 0.4.13 ( #9594 ) feat(sdk-py): add cancel_reason to Run ( #9571 ) docs(sdk-py): show the run's status in the stream example's end event ( #9588 ) feat(sdk-py): add select to threads.get_state and get_history ( #9568 )
OpenAI introduces its National Community Compact and Compute Campus Policy Blueprint to protect local resources, create jobs, and guide responsible AI infrastructure development.
创业失败后一直接外包赚钱养家，但去年和今年的行情太差了，没赚到钱，只能去找一个份工作。到今天工作时间 2 个月多一点，用 AI 写的代码。从一开始 AI 写的每一行都会 review 一下，到现在直接扔给 AI ，写完就流程体验和测试一下，不看代码，或者只看一点点主要的代码，看看改了啥东西。两个月的转变。 我发现，虽然是人指挥着 AI 在写代码，但这个过程更加累。 两方面， 一个是人脑同时对多个任务的处理是有限的（或者只有我自己），我习惯一个任务一任务进行下去，试过同时三件任务同时下达，然后等着 AI 完成后来验证，会很乱，根本不清楚上面告诉了 AI 具体是什么任务，AI 反馈要我的选择，我也不清楚到底哪个正确，只知道按 AI 的推荐确定下去。 另一方向是写了这么多年代码，面对程序出现 BUG 是一种常见的事情，但现在面对 BUG 担忧更大，心里更加没底。以前遇到 BUG 多种方式进行排查或者补救，但现在出现问题只能扔给 AI 自己去排查，然后对乱了的数据哪里乱了，需要怎么修复也是由 AI 检查后告诉我，我来运行 SQL 。但具体为啥会出现这种情况，如果避免心里没底。 之前发过一个帖子