Literature review · Pluralis · Summary brief · August 2026

WAN-PP Research Brief

The capstone of a four-part review of speculative decoding for pipeline-parallel inference over WAN. One page: where the field stands, the works that matter most for this setting, and the open gaps — each gap sized and pointed at the papers it starts from. Details, evidence, and every claim's source live in Parts I–IV.

4
levers — amortize laps · shrink the wire · fill bubbles · overlap async — all compose
5→80
tok/s headroom on a 200 ms lap, from acceptance length alone (model, Part III)
1
WAN network shipping SD (BloomBee, Apr '26) — production engines still at zero
12
open gaps below, each a publishable or buildable unit
01

State of play in five sentences

Over an N-hop WAN pipeline, decode latency is a lap of link RTTs per token, and speculative verification is the only known technique that divides the lap count — accepted length L is an RTT divisor, worth 5→25–80 tok/s on a 200 ms lap before any other optimization. The four levers are now all evidenced: amortize (trees grow acceptance ≈ log n without bound — chains cap at 1/(1−α); ~21 accepted/pass measured at 1024-node trees), shrink the wire (exact verification needs ~50 B/round, not kilobytes; 3–4-bit inter-stage activations are viable), fill bubbles (in-pipe tree schedules reach 4.5–7.8× over plain PP, with a depth wall at ~8–14 stages), and overlap (L_async = max(T_draft, T_RTT+T_verify); asynchrony beats autoregressive for any α>0). Research prototypes demonstrate every piece separately — several on real or simulated WAN links. No production engine composes speculation with layer-PP (vLLM: "not composable"; SGLang: pp_size==1 for its newest drafter); among WAN networks only BloomBee (Apr '26, a Petals descendant) ships speculation — EAGLE-2 with learned tree pruning, 1.76× over Petals at 20 Mbps — while Petals, Parallax, and exo still decode one token per lap. The field's one cautionary number: naive speculation over a real link runs at 0.44× — slower than nothing — so the wire protocol is as decisive as the drafter.

02

Most recent and notable works for the WAN-PP setting

Curated from ~120 sources across Parts I–IV: the papers a WAN-PP research effort should actually read, newest first within each role.

WorkVenueRole in a WAN-PP designKey numberCode
Intel AI-PC pipeline shardsAug '26 · arXivClosest existing precedent: TCP layer shards + SD + attention-mask rollback, tested at simulated 100 ms hops1.79× 2-node; viable at 100 ms/hop
BloomBeeApr '26 · arXivFirst WAN substrate shipping SD: Petals descendant with DP layer placement, micro-batching, lossless 46% activation compression, EAGLE-2 + learned tree pruning (−60% bytes, 96% acceptance kept), and a break-even bandwidth formula S*1.76× vs Petals @ 20 Mbps; −43% latency
Saguaro (SSD)ICLR'26Drafting through the verify window by predicting the verification outcome; geometric fan-out cache theorem = the hedge-branch economics~30% over tuned SD; ≤5× vs AR
SpecBranchICLR'26Confidence-scaled hedge branches at uncertain tokens — what to do with excess draft budget~50% fewer rollback tokens; +15% over PEARL misaligned
DFlashICML'26Block-diffusion drafter: the whole draft block in one pass — the ideal per-lap payload generator (currently pp_size==1)>6×; ≤15× on Blackwell
PicoSpec'26 · arXivThe async latency law max(T_draft, T_RTT+T_verify) + top-K downlink; measured the 0.44× naive-SD failure2.9× edge↔cloud WAN
SpecEdgeNeurIPS'25 ⭐The only real-WAN RTT sweep (14–65 ms); depth-calibration rule verify ≈ draft + RTT−22% worst case across sweep
DSSDICML'25The exact wire protocol: ids + scalars up (<50 B), full distribution down only on rejection, residual resampled locally2.4× at 50 ms RTT, lossless
SuffixDecodingNeurIPS'25 ⭐Client-side model-free drafting — zero model access, zero extra laps; SOTA on agentic trafficup to 5.3× agentic
EAGLE-3NeurIPS'25The drafter baseline to beat; head inputs live on the last PP stage; SpecForge/Speculators industrialize training2.36× measured end-to-end (SGLang)
SpecPipe / PipeDec'25 · arXivThe bubble-filling schedule: tree grows one layer per stage-hop, two-level KV — the largest reported PP gain4.46–7.79× vs plain PPpromised
PipeSpecACL-F'25The asynchrony theory: closed-form tokens/step, >AR for any α>0 — the WAN-aware version is unwritten2.54×; 5.8 J/token
Exponential races'25 · arXivShared-RNG coupling: token-ids-only exact speculation, ≈D_KL bits/token — the information-theoretic wire optimum, unshipped0 distribution bytes
DSIICLR'25Existence proof: with free comm and replicas, async speculation ≥ AR and ≥ sync SD for any drafter — the upper bound to constrain1.29–1.92× (sim)
TAH-Quant'25 · arXiv3–4-bit inter-stage activations on the PP channel — the byte floor for lever 4 (forward pass reusable at inference)4.3× e2e (training PP)
SpecExecNeurIPS'24Mega-tree construction for expensive-traversal regimes (≈ WAN cost structure): optimal-prefix trees~21 accepted/pass @1024 nodes
PipeInferSC'24The cancellation machinery validated on slow links: ID-only cancel, placeholder ordering, always-progress floor, multibuffered KV2.15× vs sync SD; best on GbE

⭐ = spotlight. Honorable mentions, in Parts II–IV: Sequoia (NeurIPS'24, optimal trees), FlowSpec ('25, segment pruning), AMUSD ('24, minimal async template), SwiftSpec ('25, tree re-rooting), MagicDec (ICLR'25, large-batch long-context), Batch-SD-Done-Right ('25, correctness invariants), CacheGen (SIGCOMM'24, KV migration for churn), Petals/Parallax/Helix (the substrates awaiting speculation).

03

The gaps: twelve open problems, sized

Theory — the cheap papers

Bandwidth-aware tree sizing. No published policy for draft length / tree size as a function of pipeline depth, RTT, and bandwidth — only empirical scans (SPD, SpecPipe) and single-link calibration rules (SpecEdge). BloomBee's break-even threshold S* is the first published piece; Sequoia's DP optimizer with t(n) rebuilt from RTT + n·s_h/BW terms is the full paper, and Part III's §04 model is its sketch (n* ≈ 16/64/256 at 10 Mbps/100 Mbps/1 Gbps).

WAN-latency-aware guarantees. DSI's "≥ everything" theorem assumes free communication and replicated targets; PipeSpec's closed forms assume zero link latency. Re-deriving either with an RTT term decides when asynchrony pays over WAN — nobody has.

Async exactness under stochastic sampling. Every async system evaluates greedy; seeded-RNG reproducibility across asynchronous replays and outcome-speculation under temperature is untreated. Blocks any lossless claim for a sampled WAN deployment.

Protocol & measurement — the cheap experiments

Shared-RNG exponential races over a real network. Token-ids-only, exactly P-distributed, ≈D_KL bits/token — published as theory, never shipped; the open question is cross-hardware float determinism on heterogeneous volunteer nodes.

Rollback microbenchmarks. Trim vs attention-mask vs multibuffered/versioned KV — seven mechanisms published (Part IV §04), zero head-to-head latency numbers. A weekend study that every WAN protocol design needs.

Energy accounting. ~62% of drafted tokens are discarded at optimal depth; AMUSD measures energy parity, PipeSpec a 3× win — no joules-per-accepted-token comparison at matched quality exists.

Compression — the unclaimed transplants

A learned bottleneck for the residual stream. FrankenSplit-style variational bottleneck + entropy coding cut split-computing bitrate 60% in vision; never applied to LLM inter-stage activations. Below 8 bits/value today only fixed quantizers (TAH-Quant) exist.

Joint compression of tree batches. Sibling nodes share prefixes and correlate strongly; every system ships tree activations node-by-node. Low-rank or shared-basis coding of the batch is unexplored.

Anchor+delta activations across decode steps. CacheGen proved adjacent-token deltas have 2.4–2.9× lower variance in KV space; nobody delta-codes the per-step hidden state crossing each PP link.

Systems — the unclaimed builds

The composed WAN stack. BloomBee (Apr '26) claimed first-mover — a Petals descendant shipping EAGLE-2 speculation with learned tree pruning, 1.76× over Petals at 20 Mbps — proving the niche is real. But it pulls two levers of four: no hedged async drafting through the RTT window, no exact ids-only wire protocol, no retrieval drafting, no cancellation machinery. The composed stack — those pieces plus DSSD wire rules, PipeInfer cancellation, and 4–8-bit activations on one substrate — is still unbuilt.

Engine plumbing for SD × PP. vLLM RFC #44697 (broadcast accepted-token counts) is the template; SGLang's MTP draft-cache transfer is roadmapped. Whoever lands it makes EAGLE-3/MTP-under-PP available to everyone.

Client-side retrieval drafting in a pipeline. Zero model access, zero extra laps, SOTA on agentic traffic — and no PP system has ever wired it in. Likely the highest value-to-effort ratio on this list.

04

The stack to build, in one paragraph

Rate-match an asynchronous drafter to the lap (verify ≈ draft + RTT), drafting client-side by retrieval where traffic repeats and with an EAGLE-3/MTP head on the last stage otherwise; spend the rest of the in-flight window on 2–4 outcome-keyed hedge branches; verify link-sized trees (n* from the Part III model — roughly 16/64/256 nodes at 10/100/1000 Mbps for an 8B in int4); let idle stages carry the tree layer-by-layer, PipeDec-style, with FlowSpec index-set pruning; ship token ids and scalars, never distributions (DSSD), quantize inter-stage activations to 4–8 bits, roll back by attention mask, cancel by run-ID with an always-progress canonical run as the floor (PipeInfer); migrate KV on churn with CacheGen. Every clause is a published, mostly open-source result — the composition is the contribution. BloomBee — Apache-2.0, Petals-lineage, already carrying EAGLE-2 and a learned pruner — is the natural substrate to extend. If the effort yields one paper first, make it the bandwidth-aware tree-sizing policy: it is the smallest gap, it quantifies the whole design, and every other piece consumes its output.