Self-evolution offers a scalable path to stronger reasoning: a pretrained language model improves itself with only minimal external supervision. Yet existing methods either depend on extensively curated or teacher-generated training data, or, when the generator runs unsupervised, reward it by a difficulty heuristic that need not improve the solver. We introduce INFUSER, an iterative co-training framework with two co-evolving roles: a Generator that drafts questions and reference golden answers from a pool of unstructured, automatically collected documents, and a Solver that improves by training on them. The solver is trained with standard correctness rewards against the generator-provided answers, while the generator is rewarded by an optimizer-aware influence score that measures whether each proposed question would actually improve the solver on the target distribution. Because this continuous, noisy influence score is poorly served by standard GRPO, we propose DuGRPO, a dual-normalized variant of GRPO, for generator training. Together, these turn the document pool into an adaptive curriculum that favors questions useful to the current solver, not just hard ones. On Qwen3-8B-Base, INFUSER outperforms strong self-evolution baselines with over 20% relative improvement on Olympiad and SuperGPQA benchmarks, and an 8B INFUSER co-evolving generator outperforms a frozen 32B thinking generator on math and coding. Ablations confirm each design choice is necessary, and two extensions, applying INFUSER to an instruction-finetuned anchor and augmenting it with rule-verifiable RLVR data, further demonstrate the flexibility and generalizability of the framework. Code is available at https://github.com/FFishy-git/INFUSER.

Abstractitations, recent work explores online RLHF that iterates between generating on-policy responses and collecting preferences [Lee et al., 2024, Park et al., 2022]. Among online The Self-Improving Alignment (SAIL) algorithmapproaches, SAIL reduces a bilevel alignment formulation addresses distribution shift by reducing a bilevelto a computationally efficient single-level surrogate and formulation of the problem to an efficient, single-reports strong empirical gains [Ding et al., 2024]. Empirically, SAIL has demonstratedisting online pipelines are largely heuristic and do not anastrong performance on this task. However, a for-lytically control the distributional shift induced by iterative mal analysis of its convergence properties has beendata collection [Chakraborty et al., 2024, Shen et al., 2024], lacking. We identify a key theoretical challenge: which has been linked to suboptimal performance in practice the standard SAIL objective function is not guar- [Sharma et al., 2024]. To address this limita-A growing line of work argues that the coupling between tion, we propose a regularized objective, SAILreward learning and policy updates is fundamentally bilevel and should be modeled as such [Chakraborty et al., 2024].RevKL, which incorporates a reverse KullbackAs a follow-up, Ding et al. [2024] reduces the bilevel align-Leibler (KL) divergence penalty to improve the optimization landscape. Our central theoretical con-ment objective to a tractable single-level surrogate and retribution is to prove that this regularized objectiveports strong empirical gains, yet it lacks formal convergence satisfies the Polyak-Lojasiewicz (PL) conditionguarantees. Related theoretical analyses in bilevel/RLHFstyle problems exist [e.g., Yang et al., 2025, Chakrabortywithin a bounded parameter space. We establish et al., 2024, Gaur et al., 2025], yet they either focus onglobal convergence guarantees, achieving a nearlinear sample complexity.

Modern LLM workflows move coordinate-indexed objects across checkpoints: steering vectors, sparse autoencoders, top-$k$ neuron sets, attribution lists, and merge alignments. This is only well posed after fixing the model's residual-stream gauge, which we show is architecture-dependent: LayerNorm residual charts have permutation gauge $S_d$ (up to a global sign flip), while RMSNorm charts with generic per-channel gain have signed-permutation gauge $B_d = S_d \ltimes \{\pm 1\}^d$. Permutation-only alignment is therefore symmetry-incomplete for RMSNorm models. We introduce sign-marginalized Hungarian matching and prove a sharp failure mode: with decorrelated coordinates, raw signed-correlation matching has a structural permutation-accuracy ceiling at the positive-sign fraction of the true gauge, which sign-marginalization removes. We then make coordinate-preserving transport, not function-level merging, the primary object: composing saved-checkpoint local $B_d$ gauges along same-base fine-tuning trajectories recovers 91.1% of cross-run coordinates at 1500 steps versus 60.3% for endpoint matching, and the gain is not explained by merely routing through the base. The recovered gauge transfers tools that permutation-only alignment breaks: TinyLlama SAE reconstruction has NMSE 0.004 under $B_d$ versus 1.08 under $S_d$; Qwen sentiment steering preserves 95.8% of its effect versus 17.2%; refusal steering reverses sign under $S_d$; coordinate-preserving merges behave the same way. The same covariance governs stateful training: signed transport of AdamW state preserves the resumed trajectory, while permutation-only state follows a different one from a functionally identical checkpoint. Finally, gauge-sweep audits show index-level interpretability claims are reproducible only relative to an explicit gauge.

Many jailbreak attacks on large language models (LLMs) rely on a common objective: making the model respond with the prefix ``Sure, here is (harmful request)''. While straightforward, this objective has two limitations: limited control over model behaviors, yielding incomplete or unrealistic jailbroken responses, and a rigid format that hinders optimization. We introduce AdvPrefix, a plug-and-play prefix-forcing objective that selects one or more model-dependent prefixes by combining two criteria: high prefilling attack success rates and low negative log-likelihood. AdvPrefix integrates seamlessly into existing jailbreak attacks to mitigate the previous limitations for free. For example, replacing GCG's default prefixes on Llama-3 improves nuanced attack success rates from 14\% to 80\%, revealing that current safety alignment fails to generalize to new prefixes. Code and selected prefixes are released.

At an event for pharmaceutical executives, biotech founders, and researchers on Tuesday, Anthropic announced Claude Science, a major new product intended to support scientific research in the same way that Claude Code supports software engineering. Like Claude Code, Claude Science can autonomously carry out meaningful work when given concise, high-level instructions, and it has access to tools that make it particularly useful for research in computational biology and drug development. Along with launching and previewing Claude Science, which is now available to all paid Claude subscribers, Anthropic also announced that it will be using the product to pursue some of its own research into drugs for rare, neglected diseases. This is not Anthropic's first foray into AI for science. In October, the company released plug-ins that help Claude make use of scientific software and databases under the heading "Claude for Life Sciences." But unlike this earlier release, Claude Science is a full-featured, standalone product. Anthropic's decision to elevate Claude Science to the same rank as Claude Code and Claude Cowork indicates that the company is taking AI's scientific applications very seriously--or at least wants to give the impression that it is.

KV cache technology, by storing key-value pairs, helps reduce the computational overhead incurred by (LLMs).

Billions of dollars are flooding into efforts to reverse aging as scientists explore ways to return cells to a younger state. But how far off are these experimental treatments? Why "reprogramming" is the buzziest approach to reversing aging right now A startup claims it broke through a bottleneck that's holding back LLMs Will Douglas Heaven Musk v. Altman week 1: Elon Musk says he was duped, warns AI could kill us all, and admits that xAI distills OpenAI's models Michelle Kim China has approved the world's first invasive brain-computer chip--here's what's next The country wants to become a global leader in brain implants. Strong government support is expected to help accelerate that process. Researchers are decoding how signals move between body and brain, with implications for how we understand and treat conditions from obesity to anxiety. In an early step towards artificial wombs, a biotech company claims it's developed a "fully artificial" chicken egg.

When you purchase through links in our articles, we may earn a small commission. Did ChatGPT give you a perfect answer? Take an AI's effective answer and turn it into a reusable template. I've written before about how to ask ChatGPT to improve a so-so prompt . But what about those times with ChatGPT, Claude, or Gemini absolutely crushes it, giving you exactly what you needed from your prompt?

In-context learning (ICL), a nonparametric learning method based on the knowledge of demonstration sets, has become a de facto standard for large language models (LLMs). The primary goal of ICL is to select valuable demonstration sets to enhance the performance of LLMs. Traditional ICL methods choose demonstration sets that share similar features with a given query. However, we have found that the performance of these traditional ICL approaches is limited on out-of-distribution (OOD) datasets, where the demonstration set and the query originate from different distributions. To ensure robust performance in OOD datasets, it is essential to learn causal representations that remain invariant between the source and target datasets. Inspired by causal representation learning, we propose causal-aware in-context learning (CCL). CCL captures the causal representations of a given dataset and selects demonstration sets that share similar causal features with the query. To achieve this, CCL employs a novel VAE-based causal representation learning technique. We demonstrate that CCL improves the OOD generalization performance of LLMs both theoretically and empirically.

'There's this deep mystery of what, actually, is this thing?': the philosopher inside Google DeepMind AI Since 2017, Iason Gabriel has worked at the tech giant, trying to anticipate - and think through - the impact of AI. But as commercial and geopolitical pressures escalate, can ethicists make any difference? In 2017, a 33-year-old political philosopher named Iason Gabriel was told by a friend that he ought to apply for a job at DeepMind, the London-based subsidiary of Google where much of its AI research was concentrated. The suggestion was not an obvious one. Gabriel was a cheerful but intense junior academic with a passion for Vipassana meditation and what his brother calls "enthusiastic" rock climbing. At the University of Oxford, where he was a fellow at St John's College, Gabriel taught courses on political theory and wrote papers on the moral contortions of "yuppie ethics" and the ethical blind spots of effective altruism. When he wasn't there, he did crisis work for the United Nations Development Programme in Sudan and Lebanon. DeepMind, meanwhile, was the world's leading AI research lab. In part, this was because it had the financial and computational backing of Google, which had bought the company in 2014 for $650m. In part, it was because DeepMind had recently shown it could put those resources to stunning use. In Seoul, in 2016, a DeepMind system called AlphaGo defeated Lee Sedol, a South Korean Go champion, in a five-game match. The victory was significant not least because of Go's legendary complexity; the game has more possible configurations than there are atoms in the universe. Thanks to the fuss around AlphaGo, Gabriel was aware of DeepMind.