{"id":"030bc3ec-9bf6-4cdf-956c-a773202b9ad7","arxiv_id":"2502.02349","paper_version":1,"verdict":"REJECT","confidence":"MODERATE","novelty_score":2.0,"correctness_risk":"high","formal_verification":"none","parameter_count":1,"one_line_summary":"A random eviction plus V-Way hybrid cache reports hit rates on four traces without any baseline comparison.","lead":"This paper proposes a cache replacement policy that combines random eviction with a V-Way style split between tag and data storage. The authors report hit rates on four benchmark traces but provide no baseline comparison, so the claimed improvements are not demonstrated.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"No baseline comparison: the reported 'up to 80.82% hit rate' is an absolute number, not evidence of improvement over any existing policy.","rationale":"The reader's weakest_assumption is exactly the load-bearing issue: the paper claims improvement without any baseline comparison. I read the full text in good faith and found no comparison table, no sensitivity analysis around the tag-to-data ratio, and no raw results or artifacts that would allow an independent check. The abstract's 'up to 80.82% hit rate' is an absolute hit rate, not a relative gain. Section III lists only RAC's own metrics, so the reader's verdict of REJECT with high correctness risk is appropriate. I also note a secondary internal inconsistency between the 'random eviction' description in Section II-A and the LRU tag eviction in Section II-B, but the missing baseline is sufficient on its own. The concrete check I propose would settle the central claim: rerun ChampSim with LRU, random replacement, and V-Way baselines on the same traces and configuration, and compare hit rates. If the baseline comparison is provided and RAC wins, the rejection can be reconsidered; as written, the evidence does not support the advertised improvements.","tokens_in":3052,"tokens_out":2165,"duration_ms":24164,"concrete_test":"Run ChampSim with the same four traces (444.namd, 445.gobmk, 473.astar, 605.mcfs) and the same 2048-set, 16-way data cache configuration under at least three baseline policies: (a) LRU, (b) random replacement, and (c) V-Way with a comparable tag/data split. Report the LLC hit rate and IPC for each policy in the same table as RAC. If RAC's hit rate is not strictly higher than all three baselines on every trace, the central claim of improvement is unsupported.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim, stated in the abstract and Section I, is that RAC improves cache hit rates over existing policies by combining random eviction with a modified V-Way design. The only evidence in Section III and Table I is a set of absolute hit rates for RAC alone (60.82%, 33.51%, 80.82%, 42.72%) with no baseline columns, no comparison policy, and no delta over LRU, random replacement, or V-Way on the same traces. The phrase 'improvements in cache hit rates up to 80.82% hit rate' is therefore an absolute value, not a measured improvement. Without a same-configuration comparison, the observed hit rates could be equal to or lower than baseline policies, which would falsify the abstract's claim. Additionally, the design description is internally mixed: Section II-A calls the scheme a random eviction policy, but Section II-B case 2 and case 4 apply LRU to the tag table when it is full. This means that even if a comparison were added, a favorable result would not cleanly attribute the gain to random eviction versus tag-table LRU. The missing baseline is the load-bearing weakness because every 'significant improvement' sentence in the evaluation depends on an implicit comparison that the paper never provides.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes Random Adaptive Cache (RAC), a cache replacement scheme that combines random eviction with a modified V-Way tag/data decoupled design, using a 32-way tag directory and a 16-way data array. The scheme is described through four fill/eviction cases. The authors evaluate RAC in ChampSim on four benchmark traces and report absolute LLC hit rates between 33.51% and 80.82%, along with IPC values, claiming significant improvements in hit rates. The central empirical claim is that RAC improves cache hit rates over existing policies.","tokens_in":3248,"tokens_out":5461,"duration_ms":56620,"significance":"The idea of combining random eviction with a V-Way-style tag/data decoupled cache is a reasonable design direction, and the stated goal of reducing the sensitivity of V-Way to tag-to-data ratio tuning is interesting. The paper clearly enumerates the four cases of its fill/eviction logic and is appropriately cautious about the modest IPC gains and the extra hardware needed for the tag directory. However, the current manuscript does not establish the central claim: no baseline policy is simulated, no comparison table is included, and no error bars or sensitivity analysis are provided. Because the headline numbers are absolute hit rates, they cannot by themselves support the word 'improvements.' If the missing baseline experiments were added and showed a real advantage over LRU, random replacement, and V-Way on the same traces under identical configurations, this design would be a modest but useful contribution to the cache replacement literature. The paper ships no code or artifacts, so reproducibility currently rests entirely on the textual description.","major_comments":[{"comment":"The abstract and Section III claim 'significant improvements' in hit rates, but Table I reports only RAC's absolute hit rates and IPC values. There is no baseline column for LRU, random replacement, or V-Way on the same traces, and no delta or normalized comparison is given. An absolute hit rate of 80.82% is not evidence of improvement; without baseline data, the central claim of the paper is unsupported.","section":"Abstract and Section III, Table I"},{"comment":"The policy description is internally mixed: Section II-A presents RAC as a random eviction policy, but Section II-B, cases 2 and 4, applies LRU to the tag table when the tag set is full. The deployed policy is therefore a hybrid of random data eviction and LRU tag eviction. The paper should either consistently describe the policy as a hybrid or provide ablations that separate the effect of random data eviction from the effect of LRU tag management; otherwise, even a favorable comparison to a baseline would not identify which design choice caused the gain.","section":"Section II-A and Section II-B"},{"comment":"The experimental methodology is underspecified and not reproducible from the manuscript alone. The authors do not state the ChampSim version, the warmup and measurement instruction counts, or the full LLC configuration beyond associativity and set count, and no code or configuration files are released. Without these details, the reported hit rates and IPC values cannot be independently verified or compared with future work.","section":"Section III"}],"minor_comments":[{"comment":"The sentence 'This corresponding valid bit for the tag entry is made invalid' should read 'The corresponding valid bit for the tag entry is made invalid.'","section":"Section II-A"},{"comment":"The fourth trace is listed as '605.mcf s-1536B' with an odd space; the official ChampSim trace name should be used consistently.","section":"Table I and Section III"},{"comment":"The bullets state 'The variation across the three different benchmarks' even though four benchmarks are listed in the table; the text should say 'four.'","section":"Section III"},{"comment":"The related-work discussion is minimal; the authors should engage more deeply with prior randomized cache designs and V-Way variants beyond the five cited references, especially since the paper claims to extend that line of work.","section":"Section I and References"},{"comment":"The text refers to Fig. 1 in several places, but the figure is not embedded in the submitted text; the camera-ready version must include it to make the pointer and eviction description understandable.","section":"Figure 1"}],"recommendation":"major_revision","confidential_remarks":"The missing baseline comparison is the central problem and is fixable within the scope of a revision. I recommend requiring the authors to rerun all four traces under at least LRU, random replacement, and V-Way baselines with identical ChampSim configurations, and to clarify or ablate the hybrid random-plus-LRU behavior. If the authors cannot provide such comparisons, the paper should not be accepted. I would not reject outright because the required experiments are well within the scope of a major revision."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Quick take: the paper describes a random/LRU hybrid cache placement scheme built on V-Way, and it is clearly written and honest about the modest IPC gains. But the central claim of 'significant improvements' is unsupported because the evaluation reports only absolute hit rates for RAC itself, with no comparison against LRU, V-Way, or random replacement on the same traces. I agree with the stress-test: 'up to 80.82% hit rate' is an absolute number, not a measured improvement.\n\nWhat's actually new is modest. The tag/data decoupling comes from V-Way [1]; randomized eviction is not new. The paper's specific combination—random eviction from the data array, LRU for the tag array when full—is a simple, plausible hybrid, and it is described clearly. It also cites the relevant prior work and does not oversell the IPC results.\n\nThe soft spots are, first, no baselines anywhere. Table I has IPC, accesses, hits, misses, and hit rate for four traces, but no column for any other policy under the same simulator and configuration. You cannot say 'significant improvements' without that. Second, the design description mixes random and LRU, so even with baselines you would not know which mechanism drives the gain. Third, no code, configuration, or ChampSim setup is shipped, making independent reproduction a real effort. The paper also does not report variance or sensitivity to the tag-to-data ratio, which it itself identifies as the key V-Way parameter.\n\nOn the positive side, the writeup is short, readable, and honest about the moderate IPC impact. The four traces are standard ChampSim traces. But absolute hit rates on four traces, with no comparison, do not support the abstract's claim.\n\nThis paper is for someone curious about yet another V-Way variant, but it is not ready for a serious referee. The authors should add baseline comparisons (at least LRU, random, and V-Way on the same configuration) and ship the simulation setup. Until then, I would desk reject rather than spend reviewer time on it.","headline":"A clearly written hybrid cache design that is undone by the absence of any baseline comparison, making the headline hit-rate claim unverifiable.","tokens_in":3773,"tokens_out":2464,"would_cite":false,"duration_ms":26909,"reading_group":"no","serious_thinker":"yes","would_accept_peer_review":false},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper claims that RAC, a hybrid of random eviction and a V-Way style tag/data split, raises last-level cache hit rates to 80.82% on an astar trace and improves hit rates across four benchmarks.","keywords":["cache replacement policy","random eviction","V-Way cache","set-associative cache","tag-data decoupling","cache hit rate","ChampSim simulation","last-level cache"],"falsifier":"Run RAC and a standard LRU policy, along with V-Way, on the same four ChampSim traces under identical cache geometry and compare last-level cache hit rates; if LRU or V-Way matches or exceeds RAC's reported hit rates, the central improvement claim fails.","tokens_in":2849,"feed_emoji":"🎯","tokens_out":6635,"duration_ms":67693,"temperature":0.7,"pith_summary":"The paper tries to establish that a cache replacement policy called RAC, for Random Adaptive Cache, improves last-level cache hit rates by combining random eviction with a V-Way style separation of tag and data storage. The design pairs a 32-way tag directory with a 16-way data array, so each set can remember more tags than it has data lines, and when the data array is full a random line is evicted to make room. On four ChampSim traces the authors report hit rates of 60.82%, 33.51%, 80.82%, and 42.72%, with the best result on the astar workload. The authors argue this approach reduces the need to hand-tune the tag-to-data ratio that V-Way caches require while keeping the flexibility of variable effective associativity. A sympathetic reader would care because better cache hit rates can cut memory access time, and a policy that adapts to uneven set demand could help real workloads without per-application tuning.","feed_headline":"Random adaptive cache policy hits 80.82% on astar trace","feed_subtitle":"Pairing random eviction with a V-Way style tag/data split lifts last-level cache hit rates on four traces.","key_machinery":"The load-bearing mechanism is the decoupled tag/data directory with forward and reverse pointers, enabling a random eviction policy that gives each set more effective associativity than its data capacity. The tag directory has 32 ways per set and the data array has 16 ways per set, so each set can carry tags for up to 32 lines while only 16 data lines are physically present; the extra tag slots act as a demand-adaptive pool. The reverse pointer is the key consistency mechanism: when a data line is randomly evicted, the reverse pointer locates its tag entry so that entry can be invalidated, keeping the tag directory and data array coherent.","core_discovery":"The central claim is that RAC, a cache replacement policy built on a decoupled tag directory and data array, achieves high last-level cache hit rates through randomized data eviction. The tag directory holds 2048 sets by 32 ways and stores only tags, each with a forward pointer to its data line, while the data array holds 2048 sets by 16 ways and stores the data lines with reverse pointers back to their tags. When a fill finds a set's data array full but its tag set not full, the policy randomly evicts a data line from that set, follows the reverse pointer to invalidate the matching tag, then installs the new tag and writes the data into the vacated line. If the tag set is also full, LRU evicts a tag entry, and when both are full LRU handles the tag set while random eviction handles the data set. The authors report hit rates of 60.82% for namd, 33.51% for gobmk, 80.82% for astar, and 42.72% for mcfs on ChampSim, and they interpret these numbers as evidence that the hybrid policy improves cache utilization and reduces memory access time, while conceding that IPC gains are moderate.","pith_inferences":["The design points to a wider trade-off space the paper does not explore: sweeping the 32:16 tag-to-data ratio would show how much tag SRAM is needed to support random adaptive placement.","A direct next test would be a baseline comparison on the same four traces against LRU, V-Way, and pure random replacement, since the paper's 'improvement' language is meaningful only against such baselines.","Because eviction is randomized, RAC may spread conflict misses differently than deterministic policies and could be studied for resilience to conflict-based cache side channels, though the paper presents no security evaluation."],"forward_implications":["RAC's reported hit rates on the four traces, especially 80.82% on astar, indicate the policy can handle workloads with non-uniform set demand without per-workload tuning.","The decoupled 32-way tag and 16-way data directories mean a hot set can retain tags for up to 32 lines while only 16 data lines exist, which should reduce the thrashing that fixed-associativity caches suffer.","By handling tag overflow with LRU and data overflow with random eviction, RAC replaces V-Way's global replacement policy with a local two-step rule that is simpler to implement.","If the hit-rate gains prove robust against baselines, the design suggests cache SRAM can be rebalanced toward tag storage without sacrificing data capacity."],"supporting_citations":[{"why":"Supplies the V-Way cache design that RAC extends: decoupled tag and data stores with demand-based associativity.","marker":"[1]"},{"why":"Provides the ChampSim simulator used for all reported hit-rate, access, miss, and IPC numbers.","marker":"[2]"}],"fun_headline_variants":["Random adaptive cache hits 80.82% on astar","Hybrid V-Way cache boosts hit rate to 80.82%","Adaptive cache policy scores 80.82% hit rate","RAC policy lifts cache hits to 80.82%","Random eviction plus V-Way yields 80.82% hits"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the reported hit rates are actually improvements over existing policies; the paper presents no baseline hit-rate numbers, so the comparison is implicit and untested.","fun_headline_variants_meta":{"raw":{"variants":["Random adaptive cache hits 80.82% on astar","Hybrid V-Way cache boosts hit rate to 80.82%","Adaptive cache policy scores 80.82% hit rate","RAC policy lifts cache hits to 80.82%","Random eviction plus V-Way yields 80.82% hits"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000183,"raw_usage":{"total_tokens":1307,"prompt_tokens":927,"completion_tokens":380,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":543,"completion_tokens_details":{"reasoning_tokens":289}},"tokens_in":543,"tokens_out":380,"duration_ms":3946,"temperature":1.0,"reasoning_tokens":289,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-09T12:26:28.817055+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run RAC and a standard LRU policy, along with V-Way, on the same four ChampSim traces under identical cache geometry and compare last-level cache hit rates; if LRU or V-Way matches or exceeds RAC's reported hit rates, the central improvement claim fails.","supporting_citations":[{"cited_title":"The V-Way cache: demand-based associativity via global replacement","cited_arxiv_id":null,"evidence_quote":"Supplies the V-Way cache design that RAC extends: decoupled tag and data stores with demand-based associativity."}],"review_version":1}