1 00:00:00,719 --> 00:00:05,279 So everyone I will be presenting recent searches for our parody violating Susy at 2 00:00:05,279 --> 00:00:10,409 the LHC. Starting with a brief introduction, our parody is a 3 00:00:10,409 --> 00:00:14,039 multiplicative quantum number. To find this equation here, you can see that it 4 00:00:14,039 --> 00:00:19,049 depends on the particles baryon number lepton number and spin. It is plus one for 5 00:00:19,049 --> 00:00:24,449 the ceremonial particles and minus one for their super symmetric partners. Our parody 6 00:00:24,449 --> 00:00:29,459 is not a symmetry of the most general local SSM Lagrangian, and here I've shown 7 00:00:29,459 --> 00:00:33,989 the parody violating portion of the super potential. And in addition to violating 8 00:00:33,989 --> 00:00:38,669 our parody these first three terms violate lepton number, and this last one violates 9 00:00:38,669 --> 00:00:44,399 baryon number. So these terms are often prevented by imposing our parody 10 00:00:44,399 --> 00:00:47,939 conservation and as a result, very odd number and phone number are still 11 00:00:47,939 --> 00:00:53,609 conserved and the proton is stable. And as a bonus Susan is is conserved so the LSP 12 00:00:53,609 --> 00:00:59,999 is neutral and stable and is therefore a dark matter candidate. However, this 13 00:00:59,999 --> 00:01:04,919 banishing of all rpv terms might be a bit heavy handed since some amount of RPC is 14 00:01:04,919 --> 00:01:09,419 fine. For example, a model only needs to conserve baryon number or left on number 15 00:01:09,419 --> 00:01:13,139 to prevent proton decay. And in general, the limits on our PV are less stringent 16 00:01:13,139 --> 00:01:17,819 when only some of the terms are nonzero. And when you allow for IPv, there could be 17 00:01:17,819 --> 00:01:22,049 a few consequences. leptons violating terms can generate light neutrino masses, 18 00:01:22,289 --> 00:01:26,909 which is pretty nice. Additionally, the LSP can carry charge indicator model 19 00:01:26,909 --> 00:01:29,969 particles and the strength of the couplings can lead to various attack 20 00:01:29,999 --> 00:01:34,349 signatures. So down here at the bottom of this cartoon showing the effects of 21 00:01:34,349 --> 00:01:38,909 turning on one of these rpp couplings on the left from the couple of years zero you 22 00:01:38,909 --> 00:01:43,229 have our purity conservation and the LP is stable. For small couplings, the LSP 23 00:01:43,229 --> 00:01:47,069 becomes long lived and it has to displace the Kaizen detector for a bit more 24 00:01:47,069 --> 00:01:50,669 moderate coupling so you have prompt decays And finally for order one couplings 25 00:01:50,669 --> 00:01:59,879 you can simply produce particles. So here is an outline for the rest of this talk. I 26 00:01:59,879 --> 00:02:04,259 will Start off by discussing some searches for the strong production of Scorpion 27 00:02:04,259 --> 00:02:07,649 gluey nose, then the electroweak production of charred Geno's and 28 00:02:07,649 --> 00:02:12,299 neutrinos. I will then show a few Long live searches and then at the end, I will 29 00:02:12,299 --> 00:02:19,829 discuss the process of setting limits on our coupling strengths. Okay, so first we 30 00:02:19,829 --> 00:02:24,569 have a CMS search for BMS visit BSN physics and events with Jetson Sims on 31 00:02:24,569 --> 00:02:29,399 leptons. This search requires either to same sized leptons, or at least three 32 00:02:29,399 --> 00:02:33,479 leptons, and additionally requires at least two jets. And this final stage is 33 00:02:33,479 --> 00:02:38,819 rarely produced in certain model, but it's pretty common in many VSM scenarios. And 34 00:02:38,819 --> 00:02:44,699 the search targets and has interpretations for a wide range of RPC and rpv signals. 35 00:02:45,299 --> 00:02:50,189 And among these are two rpv benchmark models with coil speeds. You can see the 36 00:02:50,189 --> 00:02:55,229 Fineman diagrams above. On the left, Guinan undergoes a fight party decay via 37 00:02:55,229 --> 00:03:01,229 the Qd coupling and on the right and undergoes a three body decay to the We 38 00:03:01,229 --> 00:03:07,679 have the UDP coupling into a TBS file. This search has a very ambitious single 39 00:03:07,679 --> 00:03:12,869 region strategy with over 150 orthogonal signal regions been in various kinematics 40 00:03:12,869 --> 00:03:17,699 variables. And on the left here you can see the yields in just 43 of these signal 41 00:03:17,759 --> 00:03:20,969 regions. And unfortunately, the data agrees pretty well with the background. 42 00:03:24,269 --> 00:03:27,959 And so no significant deviation we've seen from the standard model. So they set 43 00:03:27,959 --> 00:03:31,709 limits on all these various models and above you can see the results for the two 44 00:03:31,709 --> 00:03:37,169 rpv models. And you can see that Gleaner masses can be fitted up to 2.1 TV for the 45 00:03:37,169 --> 00:03:41,699 five party display on the left, and 1.7 TV for the free market decay on the right. 46 00:03:46,229 --> 00:03:50,459 Okay, next we have an atlas search in a very similar final state. The search 47 00:03:50,459 --> 00:03:55,289 requires two or three same side leptons plus jets, but the similar strategy is 48 00:03:55,289 --> 00:03:58,889 much simpler. They defined five overlapping signal regions which they fit 49 00:03:58,889 --> 00:04:03,449 independently. And among these is a dedicated rpv signal region, which 50 00:04:03,449 --> 00:04:08,279 requires exactly two Samsung leptons, at least six jets and an effective mass of 51 00:04:08,279 --> 00:04:12,959 2.6. TV for the effective mass is defined to be the scalar PT sum of all the leptons 52 00:04:12,989 --> 00:04:17,519 jets met in the event. And this region is defined pretty inclusively allowing it to 53 00:04:17,519 --> 00:04:22,379 be sensitive to many rpp scenarios. Fortunately, again knows nothing in excess 54 00:04:22,379 --> 00:04:26,819 was seen. So they said exclusions, and they have an RP v model that's similar to 55 00:04:26,819 --> 00:04:31,559 the CMS one with the Galeano decaying to a TBS and TBD final state. But instead of 56 00:04:31,559 --> 00:04:35,759 having a three body decay, they assume as intermediate stop is on shell. So they do 57 00:04:35,759 --> 00:04:40,229 a 2d mass can between the stopping masses and gluto masses can be excluded up to 58 00:04:40,229 --> 00:04:42,539 1.6. tv in this model. 59 00:04:46,410 --> 00:04:52,470 Next we have an atlas search targeting the production of gluey nose and a variety of 60 00:04:52,500 --> 00:04:57,030 pretty concerning and violating models, all of which have large jet multiplicities 61 00:04:57,060 --> 00:05:02,520 in real met and final state and The search targets the same rpv model as the Atlas 62 00:05:02,520 --> 00:05:07,500 Samsung search, you can see the Fineman diagram again on the right. And this model 63 00:05:07,500 --> 00:05:12,750 can have real met if the topic is the pinnacle to a hydronic town and atrium. 64 00:05:14,340 --> 00:05:18,540 This search performs three separate multiband fits requiring at least eight, 65 00:05:18,780 --> 00:05:23,460 nine or 10 jets with further bidding and the number of B jets and the sum of the 66 00:05:23,460 --> 00:05:28,410 masses of the larger radius jets in the event. Again, no significant excess was 67 00:05:28,410 --> 00:05:34,620 seen in Ford this rpv scenario, they can exclude the masses up to 1.5 TV and these 68 00:05:34,620 --> 00:05:42,240 contours can be compared with those from the Atlas same time search Next we have 69 00:05:42,240 --> 00:05:46,050 another Atlas search this time for bsm phenomenon and events with large Abuja 70 00:05:46,050 --> 00:05:51,240 multiple cities, no leptons and low mat and this is the first Aleksey search in 71 00:05:51,240 --> 00:05:56,370 this final state. And this search has a couple of benchmark IPv models, both 72 00:05:56,370 --> 00:06:02,100 involving stop pair production, but with different available decay paths So, on the 73 00:06:02,100 --> 00:06:05,430 left, you have this simpler model where the stock decays to a bottom insured, you 74 00:06:05,430 --> 00:06:09,150 know, and 100%. And on the right you have this other more complicated model where 75 00:06:09,150 --> 00:06:14,130 you have zero lsps. And so you have a 5050 split in between stop to bottom charge, 76 00:06:14,130 --> 00:06:20,160 you know, and stop to top, you truly know. In either case, the junos neutrinos decay 77 00:06:20,160 --> 00:06:25,620 via the UDT rpv term. This search simultaneously fits eight orthogonal 78 00:06:25,620 --> 00:06:29,160 signal regions, which are been in the number of jets and the number of digits. 79 00:06:29,430 --> 00:06:33,030 And you can see the yields in these eight signal regions on the left. And on the x 80 00:06:33,030 --> 00:06:37,110 axis, you can see the jet multiplicity requirements for each of these bins. And 81 00:06:37,110 --> 00:06:41,100 again, unfortunately, the data agrees pretty well with the expected background. 82 00:06:46,020 --> 00:06:50,250 So since no access policy limits are placed on these two models on the left for 83 00:06:50,250 --> 00:06:54,960 the simple model, and on the right for this model with things, you know, lsps and 84 00:06:54,960 --> 00:06:58,890 you can see on the right there's this diagonal across which the stop to talk to 85 00:06:58,890 --> 00:07:04,230 Kay turns on Once it becomes kinematically allowed and then these two models start 86 00:07:04,230 --> 00:07:12,420 masters can be excluded up to 950 gV. Next we have the first and only electroweak 87 00:07:12,420 --> 00:07:16,140 result this is an atlas search for trial upon residences from for Gino and 88 00:07:16,140 --> 00:07:22,380 neutralino production. This search was inspired by the B minus l MSM, where a u 89 00:07:22,380 --> 00:07:26,640 one v minus L a local gauge symmetry is added, but then spontaneously broken 90 00:07:26,910 --> 00:07:32,130 resulting in our parody and local number violation. And in this model, the Reno 91 00:07:32,130 --> 00:07:36,690 charge ginoza neutrinos are possible species. And when they are they decay 92 00:07:36,690 --> 00:07:41,040 promptly to a standard model bows on an electron or a neutrino, you can see the 93 00:07:41,040 --> 00:07:45,390 Fineman diagrams for these on the right. And the branching ratios to the different 94 00:07:45,390 --> 00:07:48,570 leptons flavors are related to the neutrino hierarchy, which is pretty 95 00:07:48,570 --> 00:07:54,930 interesting. So this search targets both see we know c one c one and C one one 96 00:07:54,930 --> 00:07:59,970 production and they require at least one charging to decay via z two, three level 97 00:08:00,870 --> 00:08:06,150 In this provides a trial upon resonance the reconstructs the tribunal mass. But 98 00:08:06,180 --> 00:08:10,320 the analysis places no constraints on the decay of the other we know they attempt to 99 00:08:10,320 --> 00:08:13,980 reconstruct the second Wiener decay as much as possible. And the number of 100 00:08:13,980 --> 00:08:17,820 leptons reconstructed bosons in the final state defines three separate signal 101 00:08:17,820 --> 00:08:23,040 regions. And on the right here you can see the MTL distribution in one of these 102 00:08:23,040 --> 00:08:27,840 signal regions. And you see the signal has nice peaks at the charging the masses. 103 00:08:30,690 --> 00:08:34,590 Unfortunately, again, I know significant access to the scene, so they set limits by 104 00:08:34,590 --> 00:08:37,620 simultaneously fitting the MTL distributions in these three signal 105 00:08:37,620 --> 00:08:42,600 regions. They then scan over the vino decay branching ratios to the different 106 00:08:42,600 --> 00:08:47,400 bosons and leptons Raiders and for each sample point in this left hand branching 107 00:08:47,400 --> 00:08:51,360 ratio space limits are set on the window mass as a function of the branching 108 00:08:51,360 --> 00:08:55,920 fraction to see as an example of one of these contours on the right under the 109 00:08:55,920 --> 00:09:01,620 hypothesis that the the you know decays are left on uniform versal As you can see, 110 00:09:01,620 --> 00:09:07,140 for this assumption, we know masses can be excluded up to 950 gV for 100%, branching 111 00:09:07,140 --> 00:09:08,160 fractions disease. 112 00:09:12,120 --> 00:09:16,530 Moving on to the Long live searches. The first is an atlas search targeting stop 113 00:09:16,530 --> 00:09:22,830 lsps that decay via the term to a muon in a downer, strange cork. And due to the 114 00:09:22,830 --> 00:09:26,790 smallness of this lambda prime coupling, the stop hydrolyzes and has a displaced a 115 00:09:26,790 --> 00:09:31,680 camera detector. So the signature for this search involves neurons with large impact 116 00:09:31,680 --> 00:09:35,850 parameters and displays vertices with transverse displacements between four and 117 00:09:35,850 --> 00:09:41,310 300 millimeters. This is before the silicon microchip tracker of Atlas. And to 118 00:09:41,310 --> 00:09:45,570 help managers to turn the model background this search vetoes any displays vertices 119 00:09:45,750 --> 00:09:49,170 that have positions consistent with the detector, either the active material or 120 00:09:49,170 --> 00:09:53,430 the supports and services. So there's really cool plot in the top right, you can 121 00:09:53,430 --> 00:09:57,960 see the positions of all these displaced vertices that fail this this material veto 122 00:09:58,500 --> 00:10:02,400 and you can see in the middle of the IV The three pixel layers and then on the 123 00:10:02,400 --> 00:10:06,780 outside you can see some support structures. Unfortunately, no events were 124 00:10:06,780 --> 00:10:10,920 seen above the expected background. And so they set limits in the stock mass versus 125 00:10:10,920 --> 00:10:16,650 lifetime plane, and stock masses can be excluded up to 1.7. tv, and these lifetime 126 00:10:16,650 --> 00:10:25,920 exclusions span four orders of magnitude. Okay, next we have a CMS search and 127 00:10:25,980 --> 00:10:31,230 targets many bsm models with longer particles decaying to displace jets. And 128 00:10:31,230 --> 00:10:35,880 among these are a few rpv models. The first is a stop model where the stock 129 00:10:35,880 --> 00:10:41,220 decays to a laptop and a downtime cork via the L Qd coupling. They also have the 130 00:10:41,220 --> 00:10:46,500 familiar Galeano to TBS model, which decays via UDP coupling with the Gleaners 131 00:10:46,500 --> 00:10:51,750 are long lived and stops rock shell. And they even have a dynamical rpv model where 132 00:10:51,750 --> 00:10:57,600 the stop decays to two down quarks. So this a synchronous analysis involves a 133 00:10:57,600 --> 00:11:01,680 digest system that's matched to a displace vertex With transverse displacements 134 00:11:01,680 --> 00:11:06,300 between our up to about 55 centimeters, this is before the outer barrel of the 135 00:11:06,300 --> 00:11:11,520 silicon strip tracker. And this search also has a material detail which you can 136 00:11:11,520 --> 00:11:15,330 see on the left you can see the beam pipe in the middle, the three pixel layers and 137 00:11:15,330 --> 00:11:19,920 then some support structures along the outside. The search expected about one 138 00:11:19,920 --> 00:11:25,530 event and saw exactly one with a displays very text that was about 26 centimeters 139 00:11:25,560 --> 00:11:29,880 out in the transverse plane, which ends up being pretty close to a silicon strip 140 00:11:29,880 --> 00:11:33,330 layer. So this was likely a standard model particle interacting with the detector. 141 00:11:34,290 --> 00:11:38,100 And in the middle here you can see the exclusion for the stock model. You can see 142 00:11:38,100 --> 00:11:48,270 the stock masses are excluded up to about 1.7 TV. CMS also developed a deep neural 143 00:11:48,270 --> 00:11:53,310 network for tagging jets from long with particle decays, and they train this Tiger 144 00:11:53,340 --> 00:11:57,750 on a split Cz model with our period of conservation. So this targeted the 145 00:11:57,750 --> 00:12:02,550 production of longer gloominess With the GUI knows the K to two courts and the 146 00:12:02,550 --> 00:12:08,490 neutralino. Then to evaluate this tiger, they calculated the expected exclusions 147 00:12:08,490 --> 00:12:13,380 for the split Susy model using the tiger. And then they compare this with a more 148 00:12:13,380 --> 00:12:18,060 conventional analysis, a few minutes of the more conventional analysis that was an 149 00:12:18,060 --> 00:12:22,170 inclusive search and final states with jets. And on this top row, you can see 150 00:12:22,170 --> 00:12:25,500 this comparison on the left for an uncompressed spectrum and on the right for 151 00:12:25,500 --> 00:12:28,860 a compressed spectrum. And you can see once you get a lot of times above about a 152 00:12:28,950 --> 00:12:34,320 millimeter, this Tiger can really help. Then to test the generalizability of this 153 00:12:34,320 --> 00:12:39,240 tiger, they evaluate its performance on GM SP and rpv models. You see the results of 154 00:12:39,240 --> 00:12:42,300 this on this bottom row on the left for short lifetime and on the right for a long 155 00:12:42,300 --> 00:12:47,700 lifetime models. And you can see both of these, these are ROC curves to show the 156 00:12:47,700 --> 00:12:50,880 mistake rate versus the tagging efficiency. And you can see this pretty 157 00:12:50,880 --> 00:12:54,600 similar performance across the models to this this Tiger seems like a generalized 158 00:12:54,600 --> 00:13:02,160 as well. Okay, so I've shown many exclusion contours for various simplified 159 00:13:02,160 --> 00:13:06,750 models. So on this slide I've organized all of them here on the right by the rpp 160 00:13:06,750 --> 00:13:11,250 coupling from the models box in red, were targeted by prompts searches, in the ones 161 00:13:11,250 --> 00:13:15,330 box in blue were targeted along with searches. So these simplified models 162 00:13:15,330 --> 00:13:20,370 assume some LSP. And some will assume the LSP as well. And they also assume some 163 00:13:20,370 --> 00:13:25,260 production mode. And then they turn on just a single rpp coupling at a time for 164 00:13:25,260 --> 00:13:29,310 the bomb searches, they assume that the coupling whatever it is, is large enough 165 00:13:29,310 --> 00:13:33,030 for properties, then they set limits on the production cross section versus the 166 00:13:33,030 --> 00:13:37,980 LSP mass, or they perform a 2d mask and well the longest search is to see what the 167 00:13:37,980 --> 00:13:42,360 coupling is small enough for displaced decay. And then they set limits on the LSP 168 00:13:42,360 --> 00:13:48,690 mass versus its lifetime. So how do these kinds of exclusions translate to limits on 169 00:13:48,690 --> 00:13:50,190 the coupling strings themselves? 170 00:13:53,760 --> 00:13:57,300 So for the longest searches, you can generally convert these lifetime limits 171 00:13:57,330 --> 00:14:02,670 into coupling limits via an equation equations can depend on other parameters, 172 00:14:02,940 --> 00:14:07,020 including the masses of any virtual particles indicate or the mixing of the 173 00:14:07,050 --> 00:14:11,640 particles. But for the problem searches, you really need reinterpretations using 174 00:14:11,640 --> 00:14:16,470 rpv signals with variable variable couplings ranks. Plus you have to add 175 00:14:16,470 --> 00:14:20,460 additional systematics for display signals since these analyses were were designed 176 00:14:20,460 --> 00:14:25,800 for prom signals and analysis targeting are pretty concerning Susie or other bsm 177 00:14:25,800 --> 00:14:29,340 physics may be sensitive to wrap up models, so we should reinterpret these as 178 00:14:29,340 --> 00:14:36,090 well. So Atlas did exactly this a couple of years ago. They reinterpreted prompt 179 00:14:36,120 --> 00:14:40,590 Suzy and exotic searches, setting limits on rpv coupling strengths in multiple 180 00:14:40,590 --> 00:14:47,640 models. Among these are is a stop model with a nonzero UDT coupling. And so you 181 00:14:47,640 --> 00:14:51,450 can see the results for the stop model on the right. And on the bottom you can see 182 00:14:51,450 --> 00:14:55,110 the relevant Fineman diagrams as you turn on the coupling, and says the same all 183 00:14:55,110 --> 00:14:59,880 from the introduction. So you can see just with these prompts searches are pretty 184 00:14:59,880 --> 00:15:04,530 good. coverage across this full range of coupling with a gap right where you'd kind 185 00:15:04,530 --> 00:15:09,330 of expect along with particle search to really help. And speaking of 186 00:15:09,330 --> 00:15:15,420 reinterpretations, both CMS and Atlas have taken steps to help facilitate, help 187 00:15:15,420 --> 00:15:19,170 facilitate them. CMS has published simplified likelihoods for their multi bit 188 00:15:19,170 --> 00:15:27,360 analyses. Analysis started publishing full likelihoods using pi h f. So summarize, I 189 00:15:27,360 --> 00:15:31,620 presented recent rpv Suzy results from Atlas and CMS targeting both short and 190 00:15:31,620 --> 00:15:36,090 long lived LSP decays. This spanned a variety of production modes mass spectrum 191 00:15:36,090 --> 00:15:39,540 and a lot of couplings resulting in a diverse set of final states. I showed 192 00:15:39,540 --> 00:15:44,160 explicit interpretations for the simplified models. But the sensitivity of 193 00:15:44,160 --> 00:15:47,280 these searches to new physics extends beyond the simplified models and even 194 00:15:47,280 --> 00:15:51,810 beyond Susy an existing analyses targeting Susie or otherwise are sensitive rpv 195 00:15:51,810 --> 00:15:55,470 Susie, so we can really benefit from more RPG interpretations. 196 00:15:57,330 --> 00:16:02,940 Thank you and take questions. From the audience so I see that Gordon has a 197 00:16:02,940 --> 00:16:06,390 question so you can unmute whenever you're ready. 198 00:16:13,920 --> 00:16:15,690 I don't think we can hear you. 199 00:16:48,030 --> 00:16:49,560 Cats. Can you hear me now? 200 00:16:50,280 --> 00:16:51,900 Yes, I can. 201 00:16:54,179 --> 00:16:59,339 Okay, so, the question I have in particular, is actually if we got back to 202 00:16:59,339 --> 00:17:06,149 the slides Aren't the rpp machine learning part? I was wondering whether we got the 203 00:17:06,149 --> 00:17:10,949 slide. So it's wondering if you've seen the third neural network, we're able to 204 00:17:10,949 --> 00:17:13,559 learn anything about the lifetimes at all. 205 00:17:14,700 --> 00:17:20,220 So the lifetimes are actually provided as an input for this tiger. So they can do 206 00:17:20,220 --> 00:17:24,480 hypothesis testing across a range of lifetimes with the single tiger. So it's 207 00:17:24,480 --> 00:17:25,470 an input for this. 208 00:17:26,430 --> 00:17:31,830 Oh, I see. So, a lifetime, I guess I imagine that when you're doing your neural 209 00:17:31,830 --> 00:17:37,800 network training, you were not using a lifetime as input. I guess I missed that. 210 00:17:38,520 --> 00:17:41,520 I know I didn't mention it. I have some more information in the backup on this. 211 00:17:41,520 --> 00:17:46,650 But yeah, they use the they use the lifetime as an input to do various 212 00:17:46,650 --> 00:17:52,530 hypothesis testing and then they train the tiger over a wide range of mass spectra 213 00:17:52,530 --> 00:17:53,400 and lifetimes. 214 00:18:02,070 --> 00:18:08,040 It's okay, thank you. Oh, and can looks like Daniel also has a question. 215 00:18:09,060 --> 00:18:13,590 Yeah, I had a question on the same slide actually. So, you know, what, what are the 216 00:18:13,590 --> 00:18:17,250 inputs that go into training this deep neural network? 217 00:18:17,969 --> 00:18:23,849 Yes, I have a slide that's kind of busy back up here, which lists all of them. So, 218 00:18:23,849 --> 00:18:28,019 they have charged a neutral Particle Flow of candidates, secondary vertices, some 219 00:18:28,019 --> 00:18:33,299 global jet features and then the lifetime as an external parameter, and the number 220 00:18:33,329 --> 00:18:37,139 of these Particle Flow candidates and the vertices and secondary vertices can vary. 221 00:18:37,289 --> 00:18:39,449 And so they can zero pad if they need to. 222 00:18:40,350 --> 00:18:45,060 Okay, um, yeah, I mean, the reason I ask is this, this really no model that you 223 00:18:45,060 --> 00:18:51,270 train on has these are hadron objects, right? They can be can pick up some 224 00:18:51,330 --> 00:18:55,800 electric charge, and it has color charge. Well, I think these gms B models don't 225 00:18:55,800 --> 00:19:01,140 have that right. The longest particle is neutral. Yes. So It's sort of surprising 226 00:19:01,140 --> 00:19:05,340 that it works for both. But is there some way to get rid of that as an input? So you 227 00:19:05,340 --> 00:19:06,990 don't have trained on that necessarily? 228 00:19:07,920 --> 00:19:12,870 Um, I'm not sure. Maybe if one of the people developing this happened to be in 229 00:19:12,870 --> 00:19:17,370 the session they can answer. But yeah, also these models have very different on 230 00:19:17,610 --> 00:19:24,270 the number of like UDS versus glue glue on vs. B and C. Cork jets varies a lot too. 231 00:19:24,270 --> 00:19:27,510 And it does pretty well across this very object composition as well. 232 00:19:29,190 --> 00:19:30,360 Great, thanks. 233 00:19:33,030 --> 00:19:38,010 All right. Thank you so much, and I think we need to move on to the next talk.