cs.CCMar 29, 2026

The Root Theorem of Context Engineering

Authors: Borja Odriozola Schick

Organizations: Independent Researcher

Read PDF
Ask AI
Cite

Copy citation

DOINot available for this paper

Abstract

Every system that maintains a large language model conversation beyond a single session faces two inescapable constraints: the context window is finite, and information quality degrades with accumulated volume. We formalize these constraints as axioms and derive a single governing principle -- the Root Theorem of Context Engineering: maximize signal-to-token ratio within bounded, lossy channels. From this principle, we derive five consequences without additional assumptions: (1)~a quality function F(P)F(P) that degrades monotonically with injected token volume, independent of window size; (2)~the independence of signal and token count as optimization variables; (3)~a necessary gate mechanism triggered by fidelity thresholds, not capacity limits; (4)~the inevitability of homeostatic persistence -- accumulate, compress, rewrite, shed -- as the only architecture that sustains understanding indefinitely; and (5)~the self-referential property that the compression mechanism operates inside the channel it compresses, requiring an external verification gate. We show that append-only systems necessarily exceed their effective window in finite time, that retrieval-augmented generation solves search but not continuity, and that the theorem's constraint structure converges with biological memory architecture through independent derivation from shared principles. Engineering proof is provided through a 60+-session persistent architecture demonstrating stable memory footprint under continuous operation -- the divergence prediction made concrete. The Root Theorem establishes context engineering as an information-theoretic discipline with formal foundations, distinct from prompt engineering in both scope and method. Shannon solved point-to-point transmission. Context engineering solves continuity.

Figures & tables

 Shannon (1948)Root Theorem
Channel typePoint-to-point transmissionSelf-referential persistence loop
Noise sourceExternal (channel interference)Internal (attention degradation)
Primary constraintBandwidthContext window + degradation
Solution strategyRedundancy (error correction)Compression (density increase)
Optimization targetBits per symbol at given error rateSignal per token at given fidelity
Channel topologySource →\to encoder →\to channel →\to decoder →\to receiverCircular: the output of one cycle becomes the input to the next
Table 1: Structural comparison between Shannon’s information theory and the Root Theorem.
Figure 1: Fidelity Degradation Curve. Three regions: Plateau (sublinear degradation, near-full fidelity), Onset (linear approximation holds), Collapse (superlinear degradation). Linear model F(P)=1−D⋅P/100KF(P)=1-D\cdot P/100\text{K} shown as reference ( D=2%D=2\% per 100K tokens). Gate position PgateP_{\text{gate}} at intersection of F(P)F(P) with FtargetF_{\text{target}} . Effective trigger fires at ∼50%{\sim}50\% of PgateP_{\text{gate}} —preserving headroom for the compression operation itself. Actual curve shape is empirical, not yet measured—the three-region structure is derived from attention mechanism behavior.
Figure 2: Divergence Chart. Memory footprint over 62 sessions. Append-only grows linearly (red); homeostatic architecture (blue) shows characteristic sawtooth pattern—growth during accumulation, reset during absorption—stabilizing around 5,000–6,500 tokens. Append-only crosses the effective context window boundary at session 10–12. After the crossing, every append-only response is generated from the degraded region. The homeostatic architecture never crosses. The sawtooth makes the regulatory mechanism visible: each downward step is a compression event.
Appendix figures & tables1 asset

Supplementary material from the paper’s appendix.

Appendix

TermDefinition
WWContext window size in tokens. Hard constraint (Axiom 1 ).
DDDegradation rate. Fractional fidelity loss per 100K tokens. Empirical ( ∼1{\sim}1 – 2%2\% /100K for frontier models).
PPPosition. Cumulative tokens consumed in the context window.
F(P)F(P)Fidelity function. Output quality at position PP : F(P)=1−(D×P/100,000)F(P)=1-(D\times P/100{,}000) .
FtargetF_{\text{target}}Target compression fidelity. System design parameter—acceptable quality loss at compression, determined by the application domain’s fidelity requirements.
PgateP_{\text{gate}}Gate position: Pgate=100,000×(1−Ftarget)/DP_{\text{gate}}=100{,}000\times(1-F_{\text{target}})/D . Independent of WW .
Appendix
Table 2: Terminology and notation.
More in this topic

Context Engineering

See all 12 papers

Explore similar work

CardsList
Aug 2, 2026

Context Compaction Theory

Large Language Models (LLMs) have a bounded context window. The context window is the maximum input size an LLM can consume for a single inference. AI agents rely on a process called context compaction to fit their state within the context window when calling an LLM. Despite its ubiquity, context compaction has received essentially no formal analysis. In this paper, we initiate a formal study of context compaction. We first introduce a framework consisting of two games that capture the two algorithmic strategies for context compaction used by contemporary AI agents in practice. The Context Selection Game models context compaction algorithms that select a subset of an agent's accumulated state to retain. The Context Generation Game models context compaction algorithms that summarize an agent's state by an arbitrary message of bounded length. We then prove an equivalence between the Context Generation Game and one-way communication complexity. The minimum context compaction budget for answering a set of queries within a target error is equal to the one-way communication complexity of the induced communication problem at the same error. Known bounds from communication complexity therefore transfer directly to context compaction. We also show that the Context Selection Game corresponds to a restricted class of one-way communication protocols. Any gap between selection and generation is therefore a gap between two classes of communication protocols. We prove that there exists a set of queries for which generation needs strictly less budget than selection. The equivalence between the Context Generation Game and one-way communication also lets us measure how well a deployed context compaction algorithm performs on a query relative to the optimal strategy. As an example, we present a case study that evaluates Anthropic's context compaction endpoint on set membership queries.

Hayder Tirmazi, Sam Markelon, Allison Bishop +1

AllSpice Inc. · Boston University · Proof Trading +2

LLM Context ManagementContext Compression
AskPDF
Jul 15, 2026

PReM: Learning What to Preserve and When to Refresh for Context Compression

Efficient long-context inference is not only about reducing memory cost, but also about keeping useful contextual evidence accessible as generation proceeds. However, existing compression-oriented approaches, such as key-value (KV) cache compression and context compression, often either make an early decision about which contextual information to keep or rely on an external compressor. Such designs make it difficult to adapt the compressed context to the evidence needed by later reasoning steps. This paper introduces PReM (Preserve and Refresh Memory), a context-compression framework that maintains the long context as the model's internal layer-wise KV memory and learns what to preserve and when to refresh it. Specifically, PReM uses a dedicated memory layer to make memory-selection decisions, and a special memory token <m> to trigger refreshes during generation. To train this behavior, PReM introduces Phase-Separated Refresh Training, aligning memory selection with memory-conditioned generation while preserving continuity across refreshes. Experiments with 32K-token contexts show that PReM outperforms strong baselines under both 16x and 32x compression, while maintaining a favorable balance between answer quality and inference efficiency.

Bohan Yu, Lei Shen, Chenxi Zhou +5

Alibaba Group

LLM Inference AccelerationLong-Context Language Model InferenceKV-Cache CompressionContext CompressionMemory-Augmented Language ModelsLLM Context Compression+1
AskPDF
May 17, 2026

Compress the Context, Keep the Commitments: A Formal Framework for Verifiable LLM Context Compression

LLM context is not just tokens; it is a set of commitments. Long-running conversations accumulate goals, constraints, decisions, preferences, tool results, retrieved evidence, artifacts, and safety boundaries that future responses must preserve. Existing context-management methods reduce length through truncation, retrieval, summarization, memory systems, or token-level prompt compression, but they rarely specify which semantic commitments must survive compression or how their preservation should be measured. We propose Context Codec, a commitment-level framework for compressing prompts and chat histories. Context Codec represents dialogue state as typed, source-grounded semantic atoms with canonical identity, equivalence, conflict, confidence, risk, and evidence spans. It separates five concerns - extraction, normalization, representation, rendering, and verification - and introduces metrics for Critical Atom Recall, Weighted Atom Recall, Commitment Density, and round-trip recoverability. It also defines a taxonomy of semantic compression errors, a concrete normalization procedure, conservative fallback rules for low-confidence and safety-critical atoms, and Context Compression Language (CCL), an ASCII-first compact rendering of canonical JSON atoms. In a small diagnostic study, CCL-Core occupies a useful middle ground between structured prose and JSON: more explicit and auditable than prose, usually more compact than JSON, and less risky than heavily minified notation. The result is not a claim that shorthand solves compression, but a framework for making context compression verifiable: compress the conversation, keep the commitments.

Natalia Trukhina, Vadim Vashkelis

Embedded Intelligence Lab (EMILAB)

LLM CompressionLLM Context CompressionLLM Context Management
AskPDF