CIS-Frame: A Multidomain Structural Framework

• CIS-Frame is a multidomain structured framework that organizes interactions, computations, and system states across diverse technical fields.
• It formalizes processes in collaborative information seeking, combinatorial games, adversarial attacks, protein geometry, V2X sensing, data streaming ISAs, and graph-theoretic analyses.
• The framework supports rigorous analysis and advances through layered models, invariant scaling laws, geometric constraints, and statically scheduled execution templates.

CIS-Frame

The term "CIS-Frame" encompasses a family of frameworks, coordinate systems, architectural patterns, or formal models unified by the acronym "CIS" within diverse technical domains: collaborative information seeking, impartial combinatorial games, backdoor attacks via continuous interaction spaces in object detection, protein geometry, V2X wireless sensing, composable instruction sets, and graph-theoretic characterization. Though differing in application, these frameworks employ the concept of a "frame"—a well-specified structure organizing interaction, computation, or system state—to yield insight, robustness, or compositionality.

1. CIS-Frame in Collaborative Information Seeking

The "CIS-Frame" formalizes true collaboration in information seeking environments, as developed by Shah (0908.0709). Collaboration is defined as a process where multiple parties, viewing different aspects of a problem, constructively resolve differences and generate solutions exceeding the sum of individual work, with final authority collectively vested.

Key System Layers and Model

• Four-Layer Extension: The canonical single-user IR model (Information, Tools, User, Results) is extended for multiple users (U):
  • Each $u_i$ traverses independent "Tool", "User", and "Result" layers; all share "Information."
  • Inter-user links include shared communication channels, shared tool interfaces, and a joint result space.
• Formal Session Tuple:

$$\mathrm{CIS} = (U, I, \mathcal{T}, \mathcal{C}, \mathcal{K}, \mathcal{A})$$

with $U$ (users), $I$ (information objects), $\mathcal{T}$ (tasks), $\mathcal{C}$ (communications), $\mathcal{K}$ (coordination/cooperation), and $\mathcal{A}$ (aggregation).

• Necessary Collaboration Conditions:
  • Diversity of opinion
  • Independence
  • Decentralization
  • Aggregation mechanism
• Evaluation Criteria: A CIS environment is judged on its explicit support for layered communication, contribution (individual artifacts), coordination (awareness/session management), cooperation (negotiation interfaces), and collaboration (joint workspaces and shared authority).

Significance

The CIS-Frame in IR is foundational for evaluating system support for group sense-making and competitive with single-user protocols only when supporting truly collective and decentralized search, negotiation, and decision making (0908.0709).

2. CIS-Frame in Impartial Combinatorial Games

The CIS-Frame, as introduced in Garrabrant–Friedman–Landsberg's analysis of Nim and general impartial games (Garrabrant et al., 2012), is the study of "cofinite induced-subgraph" (CIS) versions of games. For a game $G=(V,E)$, a CIS-G game removes a finite forbidden set $F\subset V$, yielding $G-F$ with $V' = V \setminus F$ and $E' = E \cap [V'\times V']$.

Generic Structure and Invariants

• Perturbation Family: Each CIS-G is a canonical finite perturbation of $G$'s endgame—terminal positions are determined dynamically by restriction.
• Key Theorem for Nim: Period-two scale invariance holds:

$\lim_{k \to \infty} \frac{\pi(n2^k)}{(n2^k)^2}=c$

for the number of $\mathsf{P}$-positions, even when local behavior is altered by $F$.

• Strategy Characterization: The bitwise $\oplus$ rule is destroyed, but each $(x,y)$ in Nim corresponds to a unique $z$, found recursively, such that $\{x,y,z\}$ is $\mathsf{P}$ in $\mathrm{Nim}-F$.
• Broader Implication: The CIS-Frame concept isolates large-scale, stable geometric or combinatorial properties (such as scaling laws) from artifacts introduced by finite terminal sets.

Generalization

This formalism motivates analogous CIS-analyses for other games, identifying invariants robust to local perturbations and offering a universal framework for distinguishing stable game geometries from accidental endgame effects (Garrabrant et al., 2012).

3. CIS-Frame for Continuous Interaction Space Backdoor Attacks

In adversarial machine learning, the "CIS-Frame" specifies the design and embedding of a backdoor in object detectors using continuous spatial interaction patterns instead of conventional pixel or patch-based triggers (Zhao et al., 16 Dec 2025).

Framework Construction

• Object and Interaction Encoding: Each object is parameterized as $$o_i = [c_i, x_i^{\min}, y_i^{\min}, x_i^{\max}, y_i^{\max}]$$.
• Interaction Scoring:

$$J(C_r) = \frac{N_{IoU>0}(C_s, C_r)}{N_{total}(C_s, C_r)} + \overline{IoU}(C_s, C_r)$$

identifies object-pair classes for attack salience.

• Geometric Constraints and ICS: For classes $C_s$/$C_r$, inter-object box offsets $\Delta_k$ are bounded in empirical ranges, encoding the "continuous interaction space":

$$\text{ICS} = \{(o_i,o_j): c_i=C_s, c_j=C_r,\, \bigwedge_{k=1}^N g_k(b_i,b_j)\}$$

• Sample Filtering and Label Poisoning: Only images with object pairs in ICS triggers are poisoned, supporting both single-object (OMA/ODA) and simultaneous multi-object attacks.
• Backdoor Training: Losses preserve standard detection objectives, making the attack robust to geometric and environmental variation.

Significance

This CIS-Frame generalizes the notion of "trigger" to spatial configurations—enabling invariance, multi-object coordination, and substantially enhanced attack stealth and generalization relative to traditional paradigms (Zhao et al., 16 Dec 2025).

4. CIS-Frame in Protein Geometry: The CNO-Frame

In structural biology, the CIS-Frame—also termed the CNO-Frame—serves as an intrinsic, right-handed coordinate system anchored at the carbonyl carbon ($\mathrm{C}_i$) of a peptide plane, critical for analyzing cis peptide conformations (Hou et al., 2017).

Mathematical Definition

• Axes:
  • $$\mathbf{x}_i = \frac{\vec{r}_{O_i} - \vec{r}_{C_i}}{||\vec{r}_{O_i} - \vec{r}_{C_i}||}$$
  • $$\mathbf{u}_i = \frac{\vec{r}_{N_i} - \vec{r}_{C_i}}{||\vec{r}_{N_i} - \vec{r}_{C_i}||}$$
  • $$\mathbf{z}_i = \frac{\mathbf{x}_i \times \mathbf{u}_i}{||\mathbf{x}_i \times \mathbf{u}_i||}$$
  • $\mathbf{y}_i = \mathbf{z}_i \times \mathbf{x}_i$
• Application: Projecting coordinates of neighboring atoms into this frame highlights systematic outlier deviations for cis peptide planes (Ramachandran $|\omega|<\pi/4$).
• Visualization: Displacements of atoms in CIS-Frame coordinates exceeding empirical thresholds are used to flag cis conformations and diagnose structural anomalies, enabling robust VR-based refinement.

Context

The CIS-Frame supplies an intrinsic, residue-local coordinate system for fine-grained structural analysis surpassing the diagnostic power of global or extrinsic geometric measures in protein crystallography (Hou et al., 2017).

5. CIS-Frame in Sensing-Assisted V2X Communications

In wireless networking, the CIS-Frame defines a frame structure for integrated sensing and communication that capitalizes on ISAC signals in V2I (vehicle-to-infrastructure) scenarios, enabling enhancements in pilot overhead, beam alignment, and fault tolerance (Li et al., 2023).

Structure and Protocol

• Initial Access: Replaces 64-beam SSB sweeping with radar-based omnidirectional sensing, followed by single-beam SSB transmission. Achieves up to 98.4% reduction in pilot RE overhead, reducing establishment time from 5 ms to 1.375 ms.
• Connected Mode: Eliminates periodic CSI-RS by employing sensing echoes and EKF-based tracking, reducing reference signal overhead by 43.24% and increasing throughput by 10–20% at high SNR.
• Beam-Failure Recovery: Kinematic change detection reduces failure detection time by ~60%; recovery integrates sub-6 GHz fallback or NLoS DOA extraction.
• Simulation Results:

Metric	Conventional	CIS-Frame (ISAC)
Initial Access Overhead	4.4%	0.069%
Establishment Time	5 ms	1.375 ms
Beam Failure Detection	3.75 ms	1.5 ms
Post-Recovery Throughput	700 Mbps	950 Mbps

Implication

The CIS-Frame integrates advanced radar and beam management techniques into the NR-V2X protocol, substantially reducing signaling overhead and boosting reliability for future mobile vehicular networks (Li et al., 2023).

6. CIS-Frame in Composable Instruction Sets for Data Streaming

In accelerator architecture, the CIS-Frame is a statically scheduled, resource-centric and temporally composable execution template realized within the CIS instruction set architecture (Yang et al., 2024).

Architecture and Semantics

• Spatial composability: Each instruction configures a local FSM at a resource slot. Actions are isolated per functional unit (ALU, memory, interconnect).
• Temporal composability: Nested @R (repeat) and @T (transition) instructions capture multi-level loops and fine-grained event sequencing, replacing the role of the global program counter or microcode.
• Global Control: The sequencer issues instructions to local FSMs, which run complex, streaming micro-threads in cycle-accurate lockstep.
• Instruction Encoding: Canonical 32-bit format assigns OPCODE-GROUP, SUB-OP, slot index, FSM index, and immediate fields. Example: @C configures an ALU, @S sets a memory address stream, @A atomically activates FSM groups.

App	Overhead (CIS)	Overhead (RISC-V)
DOT	+6%	+701%
1D-CONV	+4%	+1116%
2D-CONV	+0.3%	+874%

Significance

By enforcing all loop structure, address generation, and resource coordination statically and per-resource, CIS-Frame enables near-ideal PE utilization and rapid extension to new heterogeneous accelerators without global control complexity (Yang et al., 2024).

7. CIS-Frame in Claw-Free CIS Graphs

In algebraic graph theory, the CIS-Frame is a structural and algorithmic toolkit for recognizing, decomposing, and bounding the combinatorial complexity of the class of claw-free CIS (clique-intersect-stable-set) graphs (Alcón et al., 2018).

Structural Decomposition and Recognition

• Definition: $G$ is CIS iff every maximal clique intersects every maximal stable set; $G$ is claw-free if it has no induced $K_{1,3}$.
• Decomposition: Any connected, true-twin-free, claw-free CIS graph falls into exactly one of:
  1. $p K_2 + q K_1$ for $p\ge0$, $q\in\{0,1\}$.
  2. $L(K_{n,n})$ for some $n\ge1$.
  3. $L(G'\odot K_1)$, where $G'$ is triangle-free.

• Matching Criterion: A line-graph $G=L(H)$ is CIS iff $H$ is randomly internally matchable (all maximal matchings saturate every vertex of degree at least two).

• Polynomial Recognition: Each step—component and twin reduction, line-graph recognition—admits $O(n+m)$ implementation.
• Bounds: For claw-free CIS graphs, $|V(G)| < \alpha(G)\cdot \omega(G)$, in strict contrast to general CIS graphs where $|V|$ can exceed $k\cdot\alpha\cdot\omega$ for arbitrary $k$.

Open Questions

The Erdős–Hajnal property for all CIS graphs remains unresolved: whether $\max\{\alpha(G),\omega(G)\}\ge|V(G)|^{\epsilon}$ for some $\epsilon>0$ (Alcón et al., 2018).

Summary Table: Domains and Features of CIS-Frame

Context	Formalism / Frame Role	Reference
Collaborative IR	Multi-user info-seeking, aggregation, evaluation	(0908.0709)
Combinatorial Games	Cofinite perturbed games, stable geometric invariants	(Garrabrant et al., 2012)
Adversarial Object Detection	Geometric interaction space triggers	(Zhao et al., 16 Dec 2025)
Protein Geometry	Cis peptide local right-handed frame	(Hou et al., 2017)
V2X Wireless	Joint comm/sensing protocol, pilot reduction	(Li et al., 2023)
Data-Streaming ISAs	Statistically scheduled, composable instruction sets	(Yang et al., 2024)
Claw-Free CIS Graphs	Structural, recognition, and extremal bounds	(Alcón et al., 2018)

Conclusion

Across computational, biological, adversarial, and graph-theoretic disciplines, the term "CIS-Frame" designates a formal or architectural skeleton designed to clarify interaction, support rigorous analysis, enable efficient orchestration, or guarantee robustness to perturbations. Each CIS-Frame instance is anchored in explicit structure—layered interaction, induced subgraph, geometric constraint, statically scheduled microthreads, or decomposition—and is supported by precise definitions, structural theorems, or protocol descriptions. These frameworks have yielded non-trivial advances in analytical tractability, performance, recognition, and system robustness within their respective areas (0908.0709, Garrabrant et al., 2012, Zhao et al., 16 Dec 2025, Hou et al., 2017, Li et al., 2023, Yang et al., 2024, Alcón et al., 2018).