A brand new framework from Stanford College and SambaNova addresses a crucial problem in constructing strong AI brokers: context engineering. Referred to as Agentic Context Engineering (ACE), the framework robotically populates and modifies the context window of huge language mannequin (LLM) purposes by treating it as an “evolving playbook” that creates and refines methods because the agent good points expertise in its surroundings.
ACE is designed to beat key limitations of different context-engineering frameworks, stopping the mannequin’s context from degrading because it accumulates extra data. Experiments present that ACE works for each optimizing system prompts and managing an agent's reminiscence, outperforming different strategies whereas additionally being considerably extra environment friendly.
The problem of context engineering
Superior AI purposes that use LLMs largely depend on "context adaptation," or context engineering, to information their conduct. As an alternative of the pricey strategy of retraining or fine-tuning the mannequin, builders use the LLM’s in-context studying skills to information its conduct by modifying the enter prompts with particular directions, reasoning steps, or domain-specific data. This extra data is normally obtained because the agent interacts with its surroundings and gathers new knowledge and expertise. The important thing aim of context engineering is to arrange this new data in a approach that improves the mannequin’s efficiency and avoids complicated it. This strategy is changing into a central paradigm for constructing succesful, scalable, and self-improving AI methods.
Context engineering has a number of benefits for enterprise purposes. Contexts are interpretable for each customers and builders, will be up to date with new data at runtime, and will be shared throughout totally different fashions. Context engineering additionally advantages from ongoing {hardware} and software program advances, such because the rising context home windows of LLMs and environment friendly inference methods like immediate and context caching.
There are numerous automated context-engineering methods, however most of them face two key limitations. The primary is a “brevity bias,” the place immediate optimization strategies are inclined to favor concise, generic directions over complete, detailed ones. This could undermine efficiency in advanced domains.
The second, extra extreme problem is "context collapse." When an LLM is tasked with repeatedly rewriting its whole collected context, it might undergo from a type of digital amnesia.
“What we name ‘context collapse’ occurs when an AI tries to rewrite or compress every thing it has realized right into a single new model of its immediate or reminiscence,” the researchers mentioned in written feedback to VentureBeat. “Over time, that rewriting course of erases necessary particulars—like overwriting a doc so many instances that key notes disappear. In customer-facing methods, this might imply a assist agent instantly dropping consciousness of previous interactions… inflicting erratic or inconsistent conduct.”
The researchers argue that “contexts ought to operate not as concise summaries, however as complete, evolving playbooks—detailed, inclusive, and wealthy with area insights.” This strategy leans into the energy of recent LLMs, which might successfully distill relevance from lengthy and detailed contexts.
How Agentic Context Engineering (ACE) works
ACE is a framework for complete context adaptation designed for each offline duties, like system immediate optimization, and on-line situations, reminiscent of real-time reminiscence updates for brokers. Relatively than compressing data, ACE treats the context like a dynamic playbook that gathers and organizes methods over time.
The framework divides the labor throughout three specialised roles: a Generator, a Reflector, and a Curator. This modular design is impressed by “how people study—experimenting, reflecting, and consolidating—whereas avoiding the bottleneck of overloading a single mannequin with all duties,” in line with the paper.
The workflow begins with the Generator, which produces reasoning paths for enter prompts, highlighting each efficient methods and customary errors. The Reflector then analyzes these paths to extract key classes. Lastly, the Curator synthesizes these classes into compact updates and merges them into the prevailing playbook.
To forestall context collapse and brevity bias, ACE incorporates two key design ideas. First, it makes use of incremental updates. The context is represented as a group of structured, itemized bullets as a substitute of a single block of textual content. This enables ACE to make granular modifications and retrieve essentially the most related data with out rewriting your complete context.
Second, ACE makes use of a “grow-and-refine” mechanism. As new experiences are gathered, new bullets are appended to the playbook and current ones are up to date. A de-duplication step usually removes redundant entries, making certain the context stays complete but related and compact over time.
ACE in motion
The researchers evaluated ACE on two forms of duties that profit from evolving context: agent benchmarks requiring multi-turn reasoning and gear use, and domain-specific monetary evaluation benchmarks demanding specialised data. For top-stakes industries like finance, the advantages prolong past pure efficiency. Because the researchers mentioned, the framework is “way more clear: a compliance officer can actually learn what the AI realized, because it’s saved in human-readable textual content slightly than hidden in billions of parameters.”
The outcomes confirmed that ACE persistently outperformed sturdy baselines reminiscent of GEPA and basic in-context studying, attaining common efficiency good points of 10.6% on agent duties and eight.6% on domain-specific benchmarks in each offline and on-line settings.
Critically, ACE can construct efficient contexts by analyzing the suggestions from its actions and surroundings as a substitute of requiring manually labeled knowledge. The researchers observe that this potential is a "key ingredient for self-improving LLMs and brokers." On the general public AppWorld benchmark, designed to guage agentic methods, an agent utilizing ACE with a smaller open-source mannequin (DeepSeek-V3.1) matched the efficiency of the top-ranked, GPT-4.1-powered agent on common and surpassed it on the harder check set.
The takeaway for companies is critical. “This implies firms don’t should rely on large proprietary fashions to remain aggressive,” the analysis crew mentioned. “They’ll deploy native fashions, defend delicate knowledge, and nonetheless get top-tier outcomes by repeatedly refining context as a substitute of retraining weights.”
Past accuracy, ACE proved to be extremely environment friendly. It adapts to new duties with a median 86.9% decrease latency than current strategies and requires fewer steps and tokens. The researchers level out that this effectivity demonstrates that “scalable self-improvement will be achieved with each increased accuracy and decrease overhead.”
For enterprises involved about inference prices, the researchers level out that the longer contexts produced by ACE don’t translate to proportionally increased prices. Fashionable serving infrastructures are more and more optimized for long-context workloads with methods like KV cache reuse, compression, and offloading, which amortize the price of dealing with intensive context.
Finally, ACE factors towards a future the place AI methods are dynamic and repeatedly bettering. "At present, solely AI engineers can replace fashions, however context engineering opens the door for area consultants—legal professionals, analysts, medical doctors—to straight form what the AI is aware of by enhancing its contextual playbook," the researchers mentioned. This additionally makes governance extra sensible. "Selective unlearning turns into far more tractable: if a chunk of knowledge is outdated or legally delicate, it might merely be eliminated or changed within the context, with out retraining the mannequin.”
[/gpt3]
